{ "generatedAt": "2026-11-12T20:04:15Z", "brand": "Cheese Frontier", "standard": { "name": "JevCheese", "version": "jc-1.0", "method": "median_of_jevbench_axes", "rankBy": "jevbench_score_official", "axes": [ "intelligence", "calibration", "speed", "cost" ], "updateMode": "latest_tag", "fetchedAt": "2026-11-12T20:04:15Z", "sourceProtocol": "jevbench::v1.4", "sourceRevision": "v1.4.2.2", "sourceTag": "v1.4.2.2", "sourceSha256": "f0dfdd8f1601cadb16864061413e6e43c8b2dfa07b10ffd0716c67fc3c4b9952", "sourceGeneratedUtc": "2026-09-27T18:49:51+00:00", "sourceUrl": "https://raw.githubusercontent.com/fstandhartinger/jevbench/v1.4.2.2/results/v1.4.2.2/jevbench-v1.4.2.2-results.json", "sourceRepo": "https://github.com/fstandhartinger/jevbench/tree/v1.4.2.2", "liveBoard": "https://benchmarkheaven.com/jev-models", "licenseNote": "Harness & public tasks MIT; model weights/code keep their own licences.", "notes": "Auto-refreshed from the newest JevBench git tag that ships its own results JSON (GitHub Releases can lag behind tags). Board order follows official JevBench Score. JevCheese = median of I/C/S/K (Cheese Frontier lens). Cost = USD per 1,000 decisions. Attributed to Benchmark Heaven / JevBench; not merged into CF-1.3." }, "meta": { "systemCount": 95, "rankedCount": 91, "openCount": 85, "withRepoCount": 89, "openWithRepoCount": 84, "paretoCount": 8, "maxJevbenchScore": 67.37, "maxJevCheese": 88.48 }, "systems": [ { "slug": "imajev_4b", "name": "Imajev-4B", "author": "mohit67890", "repo": "https://github.com/mohit67890/imajev", "repoHost": "GitHub", "licence": "Apache-2.0 (author adapter and server metadata)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 1, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 67.37, "jevCheese": 70.04, "axes": { "intelligence": 52.2, "calibration": 80.4, "speed": 90.6, "cost": 59.7 }, "usdPer1kDecisions": 0.022, "costKind": "estimate", "value": 3182.8, "publicAccuracy": 86.1, "sealedAccuracy": 37.0, "gapPp": 49.1, "notRankedBecause": null, "rank": 1, "onPareto": true }, { "slug": "plumb-4b", "name": "Plumb-4B (crh225, JevK5 v0.2 + LoRA)", "author": "crh225", "repo": "https://github.com/crh225/plumb", "repoHost": "GitHub", "licence": "Apache-2.0 (weights and code; NOTICE credits JevK5, Qwen and SemIf); base Qwen3.5-4B Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 2, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 65.84, "jevCheese": 65.63, "axes": { "intelligence": 53.0, "calibration": 75.5, "speed": 93.5, "cost": 55.8 }, "usdPer1kDecisions": 0.0298, "costKind": "estimate", "value": 2201.6, "publicAccuracy": 89.6, "sealedAccuracy": 38.0, "gapPp": 51.6, "notRankedBecause": null, "rank": 2, "onPareto": false }, { "slug": "decider-4b-v2", "name": "decider-4b v2 (Mapika)", "author": "Mapika", "repo": "https://github.com/Mapika/decider", "repoHost": "GitHub", "licence": "Apache-2.0 (package and weights)", "class": "system-one-open", "open": "open", "underlying": null, "rankOfficial": 3, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 64.13, "jevCheese": 67.96, "axes": { "intelligence": 49.4, "calibration": 75.0, "speed": 92.9, "cost": 60.9 }, "usdPer1kDecisions": 0.0201, "costKind": "estimate", "value": 3386.5, "publicAccuracy": 83.5, "sealedAccuracy": 34.7, "gapPp": 48.8, "notRankedBecause": null, "rank": 3, "onPareto": true }, { "slug": "jev-1.13.0", "name": "Jev 1.13.0 (TypeSafe AI)", "author": "TypeSafe AI", "repo": "https://docs.typesafe.ai", "repoHost": "Docs", "licence": "proprietary API", "class": "jev", "open": "closed", "underlying": "closed", "rankOfficial": 4, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 63.29, "jevCheese": 64.7, "axes": { "intelligence": 53.1, "calibration": 76.3, "speed": 83.3, "cost": 52.0 }, "usdPer1kDecisions": 0.0399, "costKind": "measured", "value": 1621.0, "publicAccuracy": 86.6, "sealedAccuracy": 36.7, "gapPp": 49.9, "notRankedBecause": null, "rank": 4, "onPareto": false }, { "slug": "jevk5-v02", "name": "JevK5 v0.2.0", "author": "allebee", "repo": "https://github.com/allebee/jevk5", "repoHost": "GitHub", "licence": "Apache-2.0 (code/adapter); Apache-2.0 (Qwen3.5-4B base)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 5, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 62.04, "jevCheese": 67.03, "axes": { "intelligence": 48.9, "calibration": 74.5, "speed": 91.1, "cost": 59.5 }, "usdPer1kDecisions": 0.0224, "costKind": "estimate", "value": 2998.1, "publicAccuracy": 85.3, "sealedAccuracy": 33.1, "gapPp": 52.2, "notRankedBecause": null, "rank": 5, "onPareto": false }, { "slug": "cygnet", "name": "Cygnet (blockbrain, frozen Gemma-4-12B-it)", "author": "blockbrain", "repo": "https://github.com/blockbrain-ai/cygnet-recipe", "repoHost": "GitHub", "licence": "shim MIT; weights Apache-2.0 with Google's Gemma Prohibited Use Policy", "class": "system-one-open", "open": "open", "underlying": null, "rankOfficial": 6, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 61.76, "jevCheese": 63.88, "axes": { "intelligence": 49.5, "calibration": 74.9, "speed": 90.7, "cost": 52.8 }, "usdPer1kDecisions": 0.0374, "costKind": "estimate", "value": 1709.7, "publicAccuracy": 87.9, "sealedAccuracy": 33.8, "gapPp": 54.1, "notRankedBecause": null, "rank": 6, "onPareto": false }, { "slug": "hopper", "name": "Hopper", "author": "HopitAI", "repo": "https://huggingface.co/HopitAI/hopper", "repoHost": "Hugging Face", "licence": "Component-specific terms recorded in RESULT.md; submitted adapter release and Qwen base retain their respective terms", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-4B plus HopitAI/hopper LoRA", "rankOfficial": 7, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 59.43, "jevCheese": 68.89, "axes": { "intelligence": 48.0, "calibration": 79.1, "speed": 86.8, "cost": 58.7 }, "usdPer1kDecisions": 0.0238, "costKind": "estimate", "value": 2900.3, "publicAccuracy": 82.3, "sealedAccuracy": 34.1, "gapPp": 48.2, "notRankedBecause": null, "rank": 7, "onPareto": false }, { "slug": "winnow-12b", "name": "Winnow-12B Q8", "author": "Eldan Ring", "repo": "https://huggingface.co/EldanRing/Winnow-12B", "repoHost": "Hugging Face", "licence": "Apache-2.0, including the applicable Gemma 4 base/derivative licence terms", "class": "jev-rebuild", "open": "open", "underlying": "google/gemma-4-12B-it LoRA fine-tune, merged and exported as Q8_0 GGUF", "rankOfficial": 8, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 55.58, "jevCheese": 58.87, "axes": { "intelligence": 48.3, "calibration": 64.8, "speed": 82.3, "cost": 52.9 }, "usdPer1kDecisions": 0.0371, "costKind": "estimate", "value": 1587.3, "publicAccuracy": 85.7, "sealedAccuracy": 33.1, "gapPp": 52.6, "notRankedBecause": null, "rank": 8, "onPareto": false }, { "slug": "reflex-4b", "name": "reflex 4B (kshetrajna12)", "author": "kshetrajna12", "repo": "https://github.com/kshetrajna12/reflex", "repoHost": "GitHub", "licence": "MIT (code, adapter); Apache-2.0 (base)", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3.5-4B + kshetrajna12/reflex-qwen3.5-4b-lora", "rankOfficial": 9, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 53.99, "jevCheese": 63.82, "axes": { "intelligence": 47.5, "calibration": 70.4, "speed": 68.0, "cost": 59.7 }, "usdPer1kDecisions": 0.0221, "costKind": "estimate", "value": 2889.5, "publicAccuracy": 79.2, "sealedAccuracy": 28.2, "gapPp": 51.0, "notRankedBecause": null, "rank": 9, "onPareto": false }, { "slug": "djev", "name": "djev (Maisa, diffusion-gemma)", "author": "Maisa (David Villalón)", "repo": "https://github.com/Davipar/djev-dev", "repoHost": "GitHub", "licence": "Apache-2.0 code; Google DiffusionGemma Apache-2.0 weights; no djev-specific weights", "class": "jev-rebuild", "open": "open", "underlying": "inference method on google/diffusiongemma-26B-A4B-it (one structured denoising read), not a separately trained model", "rankOfficial": 10, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 52.23, "jevCheese": 56.47, "axes": { "intelligence": 47.0, "calibration": 55.4, "speed": 91.4, "cost": 57.6 }, "usdPer1kDecisions": 0.026, "costKind": "announced", "value": 2175.8, "publicAccuracy": 84.0, "sealedAccuracy": 29.9, "gapPp": 54.1, "notRankedBecause": null, "rank": 10, "onPareto": false }, { "slug": "jev-omni", "name": "Jev-Omni (akhilaaa3, Gemma-4-12B merged)", "author": "akhilaaa3", "repo": "https://huggingface.co/akhilaaa3/Jev-Omni", "repoHost": "Hugging Face", "licence": "Apache-2.0, following Gemma 4; dataset rights stated separately by the author", "class": "jev-rebuild", "open": "open", "underlying": "google/gemma-4-12B-it fine-tuned and merged, with a trained 256-way decision head", "rankOfficial": 11, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 51.34, "jevCheese": 58.55, "axes": { "intelligence": 46.8, "calibration": 64.1, "speed": 81.5, "cost": 53.0 }, "usdPer1kDecisions": 0.0368, "costKind": "estimate", "value": 1590.2, "publicAccuracy": 88.7, "sealedAccuracy": 32.1, "gapPp": 56.6, "notRankedBecause": null, "rank": 11, "onPareto": false }, { "slug": "metask-jev-4b", "name": "metask-jev-4b", "author": "Wayfind (metask-ai)", "repo": "https://github.com/metask-ai/metask-jev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-4B merged r16 LoRA, candidate-logit readout", "rankOfficial": 12, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 47.78, "jevCheese": 60.72, "axes": { "intelligence": 44.7, "calibration": 66.9, "speed": 89.1, "cost": 54.5 }, "usdPer1kDecisions": 0.0328, "costKind": "estimate", "value": 1853.5, "publicAccuracy": 79.7, "sealedAccuracy": 27.6, "gapPp": 52.1, "notRankedBecause": null, "rank": 12, "onPareto": false }, { "slug": "semif-qwen3.5-4b", "name": "SemIf, formerly OpenJev (Qwen3.5-4B, TheoLeeCJ)", "author": "Theodore Lee (TheoLeeCJ)", "repo": "https://github.com/TheoLeeCJ/openjev", "repoHost": "GitHub", "licence": "MIT (code); Qwen3.5 weights Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3.5-4B (frozen, BF16)", "rankOfficial": 13, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 47.69, "jevCheese": 63.11, "axes": { "intelligence": 44.4, "calibration": 66.8, "speed": 83.7, "cost": 59.5 }, "usdPer1kDecisions": 0.0224, "costKind": "estimate", "value": 2812.0, "publicAccuracy": 81.0, "sealedAccuracy": 26.3, "gapPp": 54.7, "notRankedBecause": null, "rank": 13, "onPareto": false }, { "slug": "jobe-qwen3.5-4b", "name": "Jobe Qwen3.5-4B (frozen)", "author": "MantisShrimpdev", "repo": "https://github.com/MantisShrimpdev/jobe", "repoHost": "GitHub", "licence": "MIT code; Apache-2.0 weights", "class": "native-logit", "open": "open", "underlying": "Qwen/Qwen3.5-4B frozen BF16 backbone", "rankOfficial": 14, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 46.94, "jevCheese": 62.81, "axes": { "intelligence": 44.1, "calibration": 66.1, "speed": 85.6, "cost": 59.5 }, "usdPer1kDecisions": 0.0224, "costKind": "estimate", "value": 2809.4, "publicAccuracy": 81.0, "sealedAccuracy": 25.6, "gapPp": 55.3, "notRankedBecause": null, "rank": 14, "onPareto": false }, { "slug": "localjev-qwen3.5-4b", "name": "local-jev Qwen3.5-4B", "author": "Amith Chandrappa (amithgc)", "repo": "https://github.com/amithgc/local-jev", "repoHost": "GitHub", "licence": "MIT code; Apache-2.0 Qwen weights", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3.5-4B text model, zero-shot next-token letter probabilities", "rankOfficial": 15, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 46.8, "jevCheese": 64.55, "axes": { "intelligence": 44.4, "calibration": 73.3, "speed": 75.0, "cost": 55.8 }, "usdPer1kDecisions": 0.0297, "costKind": "estimate", "value": 2170.0, "publicAccuracy": 80.5, "sealedAccuracy": 26.0, "gapPp": 54.5, "notRankedBecause": null, "rank": 15, "onPareto": false }, { "slug": "system-one-open", "name": "system-one-open (Gemma 4 E2B LoRA on an L4)", "author": "mithalouni", "repo": "https://github.com/mithalouni/system-one-open", "repoHost": "GitHub", "licence": "MIT (repository LICENSE; Gemma weights keep Google’s terms)", "class": "jev-rebuild", "open": "open", "underlying": "google/gemma-4-E2B-it + LoRA", "rankOfficial": 16, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 45.11, "jevCheese": 59.84, "axes": { "intelligence": 44.2, "calibration": 54.9, "speed": 77.0, "cost": 64.8 }, "usdPer1kDecisions": 0.0149, "costKind": "estimate", "value": 4021.5, "publicAccuracy": 73.2, "sealedAccuracy": 27.6, "gapPp": 45.6, "notRankedBecause": null, "rank": 16, "onPareto": true }, { "slug": "spark-s1-4b-v6", "name": "spark-s1-4b-v6 (Open Spark Jev, abhishek085)", "author": "Abhishek Rai (abhishek085)", "repo": "https://github.com/abhishek085/open-spark-jev", "repoHost": "GitHub", "licence": "Apache-2.0 (code and weights); base Qwen/Qwen3.5-4B Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3.5-4B with a LoRA, read at the option-letter logits with a fitted temperature (no extra head)", "rankOfficial": 17, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 44.62, "jevCheese": 52.73, "axes": { "intelligence": 45.1, "calibration": 47.6, "speed": 81.0, "cost": 57.9 }, "usdPer1kDecisions": 0.0253, "costKind": "estimate", "value": 2081.6, "publicAccuracy": 79.2, "sealedAccuracy": 26.6, "gapPp": 52.6, "notRankedBecause": null, "rank": 17, "onPareto": false }, { "slug": "malkuth-4b", "name": "Malkuth-4B (newfull5, Kev post-train)", "author": "newfull5 (dhtocks)", "repo": "https://github.com/newfull5/malkuth", "repoHost": "GitHub", "licence": "CC-BY-NC-4.0, research use only (XNLI and RACE in the training mix)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 18, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 44.45, "jevCheese": 61.51, "axes": { "intelligence": 43.1, "calibration": 61.5, "speed": 88.3, "cost": 61.5 }, "usdPer1kDecisions": 0.0192, "costKind": "estimate", "value": 3211.2, "publicAccuracy": 74.9, "sealedAccuracy": 23.4, "gapPp": 51.5, "notRankedBecause": null, "rank": 18, "onPareto": false }, { "slug": "jqv", "name": "jqv (Qwen3-32B zero-shot)", "author": "hjmurmur (Octalab)", "repo": "https://github.com/Octalab-Inc/jqv", "repoHost": "GitHub", "licence": "Apache-2.0 (Qwen3-32B weights); serving code public", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3-32B, bf16, read as a direct-logit classifier (no fine-tuning)", "rankOfficial": 19, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 44.35, "jevCheese": 59.55, "axes": { "intelligence": 46.4, "calibration": 71.6, "speed": 74.6, "cost": 47.5 }, "usdPer1kDecisions": 0.0564, "costKind": "estimate", "value": 1055.6, "publicAccuracy": 80.1, "sealedAccuracy": 28.2, "gapPp": 51.8, "notRankedBecause": null, "rank": 19, "onPareto": false }, { "slug": "qwen3-reranker-4b", "name": "Qwen3-Reranker-4B", "author": "Qwen", "repo": "https://huggingface.co/Qwen/Qwen3-Reranker-4B", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "reranker", "open": "open", "underlying": "4B instruction-aware generative reranker", "rankOfficial": 20, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 43.49, "jevCheese": 57.19, "axes": { "intelligence": 44.6, "calibration": 65.2, "speed": 78.7, "cost": 49.2 }, "usdPer1kDecisions": 0.0495, "costKind": "measured", "value": 1154.4, "publicAccuracy": 68.0, "sealedAccuracy": 29.9, "gapPp": 38.1, "notRankedBecause": null, "rank": 20, "onPareto": false }, { "slug": "decider-35b-a3b", "name": "decider-35b-a3b (Mapika)", "author": "Mapika", "repo": "https://huggingface.co/Mapika/decider-35b-a3b", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-35B-A3B-Base with a trained decision readout, 34.7B parameters / 3B active", "rankOfficial": 21, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 41.18, "jevCheese": 56.22, "axes": { "intelligence": 47.2, "calibration": 65.3, "speed": 80.8, "cost": 45.3 }, "usdPer1kDecisions": 0.0666, "costKind": "estimate", "value": 844.8, "publicAccuracy": 83.1, "sealedAccuracy": 31.5, "gapPp": 51.6, "notRankedBecause": null, "rank": 21, "onPareto": false }, { "slug": "raw-qwen3-4b-instruct-2507", "name": "Raw Qwen3 4B Instruct 2507 direct logits", "author": "Alibaba Qwen / neutral reproduction", "repo": "https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "raw-logit-control", "open": "open", "underlying": "Qwen3-4B-Instruct-2507 BF16", "rankOfficial": 22, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 40.95, "jevCheese": 53.04, "axes": { "intelligence": 46.4, "calibration": 29.1, "speed": 87.6, "cost": 59.7 }, "usdPer1kDecisions": 0.0221, "costKind": "estimate", "value": 2402.8, "publicAccuracy": 69.7, "sealedAccuracy": 27.3, "gapPp": 42.4, "notRankedBecause": null, "rank": 22, "onPareto": false }, { "slug": "opensourcejev-qwen35-4b-q4km", "name": "OpenSourceJev (Qwen3.5-4B Q4_K_M, native llama.cpp)", "author": "sabeel111", "repo": "https://github.com/sabeel111/OpenSourceJev", "repoHost": "GitHub", "licence": "MIT (repository code); Apache-2.0 (Qwen/Qwen3.5-4B base and unsloth/Qwen3.5-4B-GGUF Q4_K_M conversion)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 23, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 40.87, "jevCheese": 62.14, "axes": { "intelligence": 41.8, "calibration": 60.3, "speed": 82.0, "cost": 64.0 }, "usdPer1kDecisions": 0.0159, "costKind": "estimate", "value": 3907.6, "publicAccuracy": 78.4, "sealedAccuracy": 26.3, "gapPp": 52.1, "notRankedBecause": null, "rank": 23, "onPareto": false }, { "slug": "zerank-2", "name": "ZeroEntropy zerank-2", "author": "ZeroEntropy", "repo": "https://huggingface.co/zeroentropy/zerank-2-reranker", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "reranker", "open": "open", "underlying": "Qwen3-Reranker-derived 4B cross-encoder", "rankOfficial": 24, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 40.21, "jevCheese": 62.76, "axes": { "intelligence": 42.1, "calibration": 75.8, "speed": 79.0, "cost": 49.8 }, "usdPer1kDecisions": 0.0473, "costKind": "measured", "value": 1326.9, "publicAccuracy": 70.1, "sealedAccuracy": 28.6, "gapPp": 41.6, "notRankedBecause": null, "rank": 24, "onPareto": false }, { "slug": "decision-machine-1", "name": "decision-machine-1 (milliseconds.ai)", "author": "milliseconds.ai (Baptiste Laget)", "repo": "https://www.milliseconds.ai", "repoHost": "milliseconds.ai", "licence": "proprietary API, closed weights", "class": "decision-api", "open": "closed", "underlying": "decision-machine-1 (weights not published)", "rankOfficial": 25, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 39.94, "jevCheese": 61.0, "axes": { "intelligence": 41.3, "calibration": 68.3, "speed": 92.9, "cost": 53.7 }, "usdPer1kDecisions": 0.035, "costKind": "measured", "value": 1741.5, "publicAccuracy": 67.5, "sealedAccuracy": 25.6, "gapPp": 41.9, "notRankedBecause": null, "rank": 25, "onPareto": false }, { "slug": "malkuth-2b", "name": "Malkuth-2B (newfull5, Kev post-train)", "author": "newfull5 (dhtocks)", "repo": "https://github.com/newfull5/malkuth", "repoHost": "GitHub", "licence": "CC-BY-NC-4.0, research use only (XNLI and RACE in the training mix)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 26, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 38.93, "jevCheese": 57.5, "axes": { "intelligence": 41.3, "calibration": 53.5, "speed": 91.4, "cost": 61.5 }, "usdPer1kDecisions": 0.0192, "costKind": "estimate", "value": 3002.0, "publicAccuracy": 69.7, "sealedAccuracy": 24.4, "gapPp": 45.3, "notRankedBecause": null, "rank": 26, "onPareto": false }, { "slug": "raw-phi-4-mini", "name": "Raw Phi-4 mini direct logits", "author": "Microsoft / neutral reproduction", "repo": "https://huggingface.co/microsoft/Phi-4-mini-instruct", "repoHost": "Hugging Face", "licence": "MIT", "class": "raw-logit-control", "open": "open", "underlying": "Phi-4-mini-instruct BF16", "rankOfficial": 27, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 37.96, "jevCheese": 54.19, "axes": { "intelligence": 41.8, "calibration": 58.8, "speed": 88.8, "cost": 49.6 }, "usdPer1kDecisions": 0.048, "costKind": "estimate", "value": 1129.7, "publicAccuracy": 65.8, "sealedAccuracy": 29.2, "gapPp": 36.6, "notRankedBecause": null, "rank": 27, "onPareto": false }, { "slug": "jev-qwen3.5-9b-base-nvfp4", "name": "JEV Qwen3.5-9B Base NVFP4", "author": "WilfLin", "repo": "https://huggingface.co/WIlfLin/JEV-Qwen3.5-9B-Base-NVFP4", "repoHost": "Hugging Face", "licence": "Apache-2.0 entrant and upstream checkpoint", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-9B Base NVFP4 with compact BF16 decision head", "rankOfficial": 28, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 37.7, "jevCheese": 57.26, "axes": { "intelligence": 46.8, "calibration": 67.7, "speed": 93.3, "cost": 43.3 }, "usdPer1kDecisions": 0.0775, "costKind": "estimate", "value": 739.2, "publicAccuracy": 75.3, "sealedAccuracy": 29.5, "gapPp": 45.8, "notRankedBecause": null, "rank": 28, "onPareto": false }, { "slug": "openjev-razorback16", "name": "OpenJev (DiffusionGemma 26B-A4B NVFP4, razorback16)", "author": "razorback16 / Codiv", "repo": "https://github.com/razorback16/openjev", "repoHost": "GitHub", "licence": "Apache-2.0 (repo and weights)", "class": "jev-rebuild", "open": "open", "underlying": "nvidia/diffusiongemma-26B-A4B-it-NVFP4", "rankOfficial": 29, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 36.85, "jevCheese": 50.24, "axes": { "intelligence": 45.4, "calibration": 55.0, "speed": 83.2, "cost": 45.5 }, "usdPer1kDecisions": 0.0656, "costKind": "estimate", "value": 765.8, "publicAccuracy": 81.8, "sealedAccuracy": 28.6, "gapPp": 53.2, "notRankedBecause": null, "rank": 29, "onPareto": false }, { "slug": "kev-4b", "name": "kev 4B (research preview)", "author": "Jared Palmer", "repo": "https://github.com/jaredpalmer/kev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3-4B-Base + LoRA + learned pointer head; jaredpalmer/kev-4b", "rankOfficial": 30, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 36.14, "jevCheese": 51.92, "axes": { "intelligence": 42.1, "calibration": 39.6, "speed": 75.7, "cost": 61.8 }, "usdPer1kDecisions": 0.0188, "costKind": "estimate", "value": 2761.5, "publicAccuracy": 66.2, "sealedAccuracy": 22.4, "gapPp": 43.8, "notRankedBecause": null, "rank": 30, "onPareto": false }, { "slug": "decision-2b", "name": "Decision 2B (FlyMy.AI, v59)", "author": "FlyMy.AI (@denti)", "repo": "https://huggingface.co/flymy-ai/decision-2b-preview", "repoHost": "Hugging Face", "licence": "Apache-2.0 notices on the included code and the pinned base; the weights are an evaluation preview under EVALUATION-PERMISSION.md, not a cleared commercial release", "class": "jev-rebuild", "open": "open", "underlying": "openbmb/MiniCPM5-2B with a trained LoRA adapter and pointer head (26.2M trainable parameters)", "rankOfficial": 31, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 35.8, "jevCheese": 68.32, "axes": { "intelligence": 38.8, "calibration": 74.1, "speed": 84.3, "cost": 62.5 }, "usdPer1kDecisions": 0.0178, "costKind": "estimate", "value": 3844.5, "publicAccuracy": 75.3, "sealedAccuracy": 26.0, "gapPp": 49.4, "notRankedBecause": null, "rank": 31, "onPareto": false }, { "slug": "qwen35-9b-jev-data-mix-v2", "name": "Qwen3.5-9B Jev-like data-mix v2", "author": "jsaurabh", "repo": "https://huggingface.co/jsaurabh/qwen3.5-9b-jev-data-mix-v2", "repoHost": "Hugging Face", "licence": "Apache-2.0 (adapter code and weights; Qwen3.5-9B base is Apache-2.0)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 32, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 35.24, "jevCheese": 54.37, "axes": { "intelligence": 47.4, "calibration": 61.3, "speed": 82.0, "cost": 42.4 }, "usdPer1kDecisions": 0.0834, "costKind": "estimate", "value": 651.6, "publicAccuracy": 78.4, "sealedAccuracy": 29.2, "gapPp": 49.1, "notRankedBecause": null, "rank": 32, "onPareto": false }, { "slug": "gpt-6-luna-low", "name": "GPT-6 Luna (low reasoning effort)", "author": "OpenAI", "repo": null, "repoHost": null, "licence": "proprietary API", "class": "llm-baseline", "open": "closed", "underlying": null, "rankOfficial": 33, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 35.11, "jevCheese": 82.82, "axes": { "intelligence": 95.8, "calibration": 92.0, "speed": 73.7, "cost": 36.9 }, "usdPer1kDecisions": 0.127, "costKind": "measured", "value": 652.1, "publicAccuracy": 99.1, "sealedAccuracy": 92.9, "gapPp": 6.3, "notRankedBecause": null, "rank": 33, "onPareto": false }, { "slug": "simplejev-qwen3.8-27b", "name": "SimpleJev Qwen3.8-27B", "author": "Featherless AI", "repo": "https://github.com/featherless-ai/simple-jev", "repoHost": "GitHub", "licence": "Apache-2.0 (Qwen weights); repository licence not stated", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.8-27B through SimpleJev's direct-logit classifier", "rankOfficial": 34, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 34.57, "jevCheese": 61.38, "axes": { "intelligence": 51.6, "calibration": 74.5, "speed": 71.2, "cost": 39.5 }, "usdPer1kDecisions": 0.104, "costKind": "estimate", "value": 590.2, "publicAccuracy": 86.6, "sealedAccuracy": 35.7, "gapPp": 50.9, "notRankedBecause": null, "rank": 34, "onPareto": false }, { "slug": "ninfer-qwen3.8-flash-next", "name": "NInfer Qwen3.8-Flash-Next mixed", "author": "Igor L. / NInfer contributors", "repo": "https://github.com/igorls/ninfer", "repoHost": "GitHub", "licence": "Qwen Community License 1.0 (model); Apache-2.0 (engine)", "class": "native-logit", "open": "open", "underlying": "Qwen/Qwen3.8-Flash-Next, mixed NVFP4/FP8/INT4 NInfer artifact", "rankOfficial": 35, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 33.98, "jevCheese": 64.05, "axes": { "intelligence": 49.5, "calibration": 78.6, "speed": 88.2, "cost": 38.9 }, "usdPer1kDecisions": 0.1087, "costKind": "estimate", "value": 589.3, "publicAccuracy": 89.6, "sealedAccuracy": 34.1, "gapPp": 55.5, "notRankedBecause": null, "rank": 35, "onPareto": false }, { "slug": "swanone", "name": "swanOne (blockbrain, Qwen3.8-Flash-Next NVFP4)", "author": "blockbrain", "repo": "https://github.com/blockbrain-ai/swanone-recipe", "repoHost": "GitHub", "licence": "patches Apache-2.0, shim MIT; weights under the Qwen licence (LICENSE-NOTICE.md)", "class": "system-one-open", "open": "open", "underlying": null, "rankOfficial": 36, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 33.61, "jevCheese": 61.81, "axes": { "intelligence": 52.6, "calibration": 71.0, "speed": 82.5, "cost": 38.6 }, "usdPer1kDecisions": 0.111, "costKind": "estimate", "value": 556.9, "publicAccuracy": 88.7, "sealedAccuracy": 38.6, "gapPp": 50.1, "notRankedBecause": null, "rank": 36, "onPareto": false }, { "slug": "gpt-6-luna", "name": "GPT-6 Luna (default medium reasoning effort)", "author": "OpenAI", "repo": null, "repoHost": null, "licence": "proprietary API", "class": "llm-baseline", "open": "closed", "underlying": null, "rankOfficial": 37, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 33.27, "jevCheese": 83.02, "axes": { "intelligence": 97.4, "calibration": 93.5, "speed": 72.6, "cost": 36.0 }, "usdPer1kDecisions": 0.1354, "costKind": "measured", "value": 613.0, "publicAccuracy": 99.6, "sealedAccuracy": 95.5, "gapPp": 4.1, "notRankedBecause": null, "rank": 37, "onPareto": false }, { "slug": "open-alternative-jev", "name": "open-alternative-jev (Qwen3.5-4B, IkerMoel)", "author": "IkerMoel", "repo": "https://github.com/ikermoel/open-alternative-jev", "repoHost": "GitHub", "licence": "Apache-2.0 (code and weights)", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3.5-4B (frozen, BF16)", "rankOfficial": 38, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 33.24, "jevCheese": 59.16, "axes": { "intelligence": 38.6, "calibration": 58.7, "speed": 83.5, "cost": 59.6 }, "usdPer1kDecisions": 0.0222, "costKind": "estimate", "value": 2668.2, "publicAccuracy": 74.0, "sealedAccuracy": 24.4, "gapPp": 49.7, "notRankedBecause": null, "rank": 38, "onPareto": false }, { "slug": "jev-local", "name": "jev-local (Qwen3.5-9B)", "author": "us (GitHub)", "repo": "https://github.com/us/jev-local", "repoHost": "GitHub", "licence": "no licence stated in the repository (public code); Apache-2.0 base weights", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3.5-9B, frozen, per-option mean log-probability", "rankOfficial": 39, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 32.54, "jevCheese": 54.66, "axes": { "intelligence": 45.2, "calibration": 64.2, "speed": 69.2, "cost": 43.3 }, "usdPer1kDecisions": 0.0775, "costKind": "estimate", "value": 705.7, "publicAccuracy": 74.9, "sealedAccuracy": 29.5, "gapPp": 45.3, "notRankedBecause": null, "rank": 39, "onPareto": false }, { "slug": "decision-fast", "name": "Decision Fast (FlyMy.AI, v53a)", "author": "FlyMy.AI (@denti)", "repo": "https://huggingface.co/flymy-ai/decision-fast-preview", "repoHost": "Hugging Face", "licence": "Apache-2.0 notices on the included code and the pinned base; the weights are an evaluation preview under EVALUATION-PERMISSION.md, not a cleared commercial release", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3-0.6B-Base with a trained LoRA adapter and pointer head (10.6M trainable parameters)", "rankOfficial": 40, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 32.49, "jevCheese": 70.69, "axes": { "intelligence": 37.1, "calibration": 65.3, "speed": 81.6, "cost": 76.1 }, "usdPer1kDecisions": 0.0063, "costKind": "estimate", "value": 11270.2, "publicAccuracy": 63.2, "sealedAccuracy": 25.6, "gapPp": 37.6, "notRankedBecause": null, "rank": 40, "onPareto": true }, { "slug": "decider-2b", "name": "decider-2b (Mapika)", "author": "Mapika", "repo": "https://huggingface.co/Mapika/decider-2b", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-2B-Base with a trained decision readout, 1.9B", "rankOfficial": 41, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 30.74, "jevCheese": 52.24, "axes": { "intelligence": 38.5, "calibration": 43.5, "speed": 83.2, "cost": 61.0 }, "usdPer1kDecisions": 0.02, "costKind": "estimate", "value": 2616.5, "publicAccuracy": 71.0, "sealedAccuracy": 24.7, "gapPp": 46.3, "notRankedBecause": null, "rank": 41, "onPareto": false }, { "slug": "jeff", "name": "jeff (Logan Markewich, GLiFormer 400M)", "author": "Logan Markewich", "repo": "https://github.com/logan-markewich/jeff", "repoHost": "GitHub", "licence": "MIT (code); GLiFormer weights per their model card", "class": "jev-rebuild", "open": "open", "underlying": "GLiFormer large (knowledgator/gliformer-large-v1, ~400M) behind a TypeSafe-compatible /v1/systemone server", "rankOfficial": 42, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 30.58, "jevCheese": 65.69, "axes": { "intelligence": 36.8, "calibration": 67.9, "speed": 63.5, "cost": 76.6 }, "usdPer1kDecisions": 0.006, "costKind": "estimate", "value": 10880.9, "publicAccuracy": 62.8, "sealedAccuracy": 33.1, "gapPp": 29.7, "notRankedBecause": null, "rank": 42, "onPareto": true }, { "slug": "laya", "name": "Laya (Convai Innovations, ModernBERT-large 421M)", "author": "Convai Innovations", "repo": "https://huggingface.co/convaiinnovations/laya", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "ModernBERT-large encoder + option-marker decision head, 421M, RLCD-trained", "rankOfficial": 43, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 30.25, "jevCheese": 67.37, "axes": { "intelligence": 36.1, "calibration": 63.7, "speed": 71.1, "cost": 86.2 }, "usdPer1kDecisions": 0.0029, "costKind": "estimate", "value": 23366.2, "publicAccuracy": 58.4, "sealedAccuracy": 30.8, "gapPp": 27.6, "notRankedBecause": null, "rank": 43, "onPareto": true }, { "slug": "kushal-gemma4-31b-it-autoloops", "name": "Autoloops – Gemma 4 31B IT", "author": "Autoloops", "repo": null, "repoHost": null, "licence": "Gemma terms of use (Google) for the base weights; hosted Autoloops service", "class": "jev", "open": "closed", "underlying": null, "rankOfficial": 44, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 29.96, "jevCheese": 69.51, "axes": { "intelligence": 59.8, "calibration": 79.2, "speed": 84.0, "cost": 36.0 }, "usdPer1kDecisions": 0.1363, "costKind": "measured", "value": 509.8, "publicAccuracy": 92.8, "sealedAccuracy": 45.8, "gapPp": 47.1, "notRankedBecause": null, "rank": 44, "onPareto": false }, { "slug": "standardone-8b", "name": "Standard One 8B (Standard Thinking)", "author": "Standard Thinking (myeongho12)", "repo": "https://huggingface.co/StandardThinking/StandardOne-8B", "repoHost": "Hugging Face", "licence": "Apache-2.0 (jev-adapter server, LoRA and merged weights; base Ministral 3 8B Apache-2.0)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 45, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 29.07, "jevCheese": 56.02, "axes": { "intelligence": 46.2, "calibration": 65.9, "speed": 92.0, "cost": 39.4 }, "usdPer1kDecisions": 0.1044, "costKind": "estimate", "value": 536.4, "publicAccuracy": 76.6, "sealedAccuracy": 26.6, "gapPp": 50.0, "notRankedBecause": null, "rank": 45, "onPareto": false }, { "slug": "lev-350m", "name": "lev-350m (Franck Verrot, LFM2.5-350M)", "author": "Franck Verrot (franckverrot)", "repo": "https://github.com/franckverrot/lev", "repoHost": "GitHub", "licence": "Apache-2.0 (code); weights under LiquidAI's LFM1.0 licence, following the LFM2.5-350M base", "class": "jev-rebuild", "open": "open", "underlying": "LiquidAI/LFM2.5-350M with a LoRA and a 6.5M-parameter pointer head (a kev clone)", "rankOfficial": 46, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 28.5, "jevCheese": 73.35, "axes": { "intelligence": 34.8, "calibration": 70.6, "speed": 85.3, "cost": 76.1 }, "usdPer1kDecisions": 0.0063, "costKind": "estimate", "value": 11711.4, "publicAccuracy": 58.4, "sealedAccuracy": 25.0, "gapPp": 33.4, "notRankedBecause": null, "rank": 46, "onPareto": false }, { "slug": "openjev-sglang", "name": "openjev-sglang (Qwen3.6-35B-A3B on SGLang)", "author": "ekzhang", "repo": "https://github.com/ekzhang/openjev-sglang", "repoHost": "GitHub", "licence": "no licence file in the repository as of 2026-09-19; Qwen3.6 weights keep their own terms", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.6-35B-A3B", "rankOfficial": 47, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 27.65, "jevCheese": 59.33, "axes": { "intelligence": 49.4, "calibration": 69.2, "speed": 77.1, "cost": 36.5 }, "usdPer1kDecisions": 0.1313, "costKind": "estimate", "value": 451.9, "publicAccuracy": 85.3, "sealedAccuracy": 33.1, "gapPp": 52.2, "notRankedBecause": null, "rank": 47, "onPareto": false }, { "slug": "von-395m", "name": "Von (wfzyx, Option-Marker 395M)", "author": "wfzyx (Victor Hugo)", "repo": "https://github.com/wfzyx/von", "repoHost": "GitHub", "licence": "Apache-2.0 code; Apache-2.0 weights (answerdotai/ModernBERT-large base)", "class": "jev-rebuild", "open": "open", "underlying": null, "rankOfficial": 48, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 27.48, "jevCheese": 73.08, "axes": { "intelligence": 34.5, "calibration": 75.7, "speed": 70.5, "cost": 77.8 }, "usdPer1kDecisions": 0.0055, "costKind": "estimate", "value": 13256.2, "publicAccuracy": 57.1, "sealedAccuracy": 27.9, "gapPp": 29.2, "notRankedBecause": null, "rank": 48, "onPareto": false }, { "slug": "ninfer-qwen3.8-27b-t1.5", "name": "NInfer Qwen3.8-27B NVFP4 (T=1.5)", "author": "Igor L. / NInfer contributors", "repo": "https://github.com/igorls/ninfer", "repoHost": "GitHub", "licence": "Apache-2.0 (engine and submitted 27B artifact/base)", "class": "native-logit", "open": "open", "underlying": "Qwen/Qwen3.8-27B, NInfer NVFP4/FP8 artifact", "rankOfficial": 49, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 26.92, "jevCheese": 63.71, "axes": { "intelligence": 51.5, "calibration": 76.0, "speed": 80.1, "cost": 35.2 }, "usdPer1kDecisions": 0.1449, "costKind": "estimate", "value": 439.6, "publicAccuracy": 83.1, "sealedAccuracy": 33.1, "gapPp": 50.0, "notRankedBecause": null, "rank": 49, "onPareto": false }, { "slug": "ninfer-qwen3.8-27b", "name": "NInfer Qwen3.8-27B NVFP4", "author": "Igor L. / NInfer contributors", "repo": "https://github.com/igorls/ninfer", "repoHost": "GitHub", "licence": "Apache-2.0 (engine and submitted 27B artifact/base)", "class": "native-logit", "open": "open", "underlying": "Qwen/Qwen3.8-27B, NInfer NVFP4/FP8 artifact", "rankOfficial": 50, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 26.3, "jevCheese": 59.32, "axes": { "intelligence": 51.5, "calibration": 67.2, "speed": 80.1, "cost": 35.2 }, "usdPer1kDecisions": 0.1449, "costKind": "estimate", "value": 409.3, "publicAccuracy": 83.1, "sealedAccuracy": 33.1, "gapPp": 50.0, "notRankedBecause": null, "rank": 50, "onPareto": false }, { "slug": "kev-8b", "name": "kev 8B (research preview)", "author": "Jared Palmer", "repo": "https://github.com/jaredpalmer/kev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3-8B-Base + LoRA + learned pointer head; jaredpalmer/kev-8b", "rankOfficial": 51, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 25.56, "jevCheese": 42.92, "axes": { "intelligence": 41.8, "calibration": 40.2, "speed": 74.9, "cost": 44.0 }, "usdPer1kDecisions": 0.0733, "costKind": "estimate", "value": 585.4, "publicAccuracy": 71.4, "sealedAccuracy": 21.8, "gapPp": 49.7, "notRankedBecause": null, "rank": 51, "onPareto": false }, { "slug": "jevone", "name": "JevOne", "author": "Juspay", "repo": "https://huggingface.co/juspay/jev-one", "repoHost": "Hugging Face", "licence": "Component-specific JevOne/Qwen/SGLang terms recorded in RESULT.md", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.6-35B-A3B BF16 with JevOne bidirectional option-logit mapping", "rankOfficial": 52, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 25.51, "jevCheese": 62.59, "axes": { "intelligence": 47.6, "calibration": 77.6, "speed": 88.5, "cost": 35.9 }, "usdPer1kDecisions": 0.1373, "costKind": "estimate", "value": 455.8, "publicAccuracy": 89.6, "sealedAccuracy": 33.8, "gapPp": 55.8, "notRankedBecause": null, "rank": 52, "onPareto": false }, { "slug": "typecastlm", "name": "typecastlm (Mikhail Gribov, Qwen3.5-4B computed head)", "author": "Mikhail Gribov", "repo": "https://github.com/mihail-gribov/typecastlm", "repoHost": "GitHub", "licence": "Apache-2.0 (package and weights)", "class": "system-one-open", "open": "open", "underlying": null, "rankOfficial": 53, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 25.28, "jevCheese": 61.13, "axes": { "intelligence": 34.0, "calibration": 62.1, "speed": 92.6, "cost": 60.2 }, "usdPer1kDecisions": 0.0213, "costKind": "estimate", "value": 2874.1, "publicAccuracy": 77.9, "sealedAccuracy": 29.9, "gapPp": 48.1, "notRankedBecause": null, "rank": 53, "onPareto": false }, { "slug": "simplejev-qwen3.6-35b-a3b", "name": "SimpleJev Qwen3.6-35B-A3B", "author": "Featherless AI", "repo": "https://github.com/featherless-ai/simple-jev", "repoHost": "GitHub", "licence": "Apache-2.0 (Qwen weights); repository licence not stated", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.6-35B-A3B through SimpleJev's direct-logit classifier", "rankOfficial": 54, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 24.86, "jevCheese": 52.76, "axes": { "intelligence": 45.7, "calibration": 59.8, "speed": 75.0, "cost": 38.1 }, "usdPer1kDecisions": 0.1156, "costKind": "estimate", "value": 456.6, "publicAccuracy": 81.4, "sealedAccuracy": 28.2, "gapPp": 53.1, "notRankedBecause": null, "rank": 54, "onPareto": false }, { "slug": "kev-0.6b", "name": "kev 0.6B (research preview)", "author": "Jared Palmer", "repo": "https://github.com/jaredpalmer/kev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3-0.6B-Base + LoRA + learned pointer head; jaredpalmer/kev-0.6b", "rankOfficial": 55, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 24.75, "jevCheese": 62.76, "axes": { "intelligence": 34.2, "calibration": 50.0, "speed": 75.6, "cost": 76.1 }, "usdPer1kDecisions": 0.0063, "costKind": "estimate", "value": 10012.9, "publicAccuracy": 66.7, "sealedAccuracy": 24.0, "gapPp": 42.6, "notRankedBecause": null, "rank": 55, "onPareto": false }, { "slug": "raw-qwen3-8b", "name": "Raw Qwen3 8B direct logits", "author": "Alibaba Qwen / neutral reproduction", "repo": "https://huggingface.co/Qwen/Qwen3-8B", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "raw-logit-control", "open": "open", "underlying": "Qwen3-8B BF16", "rankOfficial": 56, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 23.68, "jevCheese": 43.77, "axes": { "intelligence": 45.7, "calibration": 24.1, "speed": 86.3, "cost": 41.9 }, "usdPer1kDecisions": 0.0866, "costKind": "estimate", "value": 505.7, "publicAccuracy": 68.4, "sealedAccuracy": 26.3, "gapPp": 42.1, "notRankedBecause": null, "rank": 56, "onPareto": false }, { "slug": "system-one-sg", "name": "system-one (Qwen3-8B, Sean Goedecke)", "author": "Sean Goedecke", "repo": "https://github.com/sgoedecke/system-one", "repoHost": "GitHub", "licence": "no licence file in the repository as of 19 Sep; Qwen3 weights Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3-8B (frozen, BF16), the model of the author's demos", "rankOfficial": 57, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 23.37, "jevCheese": 42.52, "axes": { "intelligence": 43.6, "calibration": 32.8, "speed": 84.4, "cost": 41.5 }, "usdPer1kDecisions": 0.0894, "costKind": "estimate", "value": 475.4, "publicAccuracy": 71.9, "sealedAccuracy": 24.4, "gapPp": 47.5, "notRankedBecause": null, "rank": 57, "onPareto": false }, { "slug": "opendecision", "name": "OpenDecision (ModernBERT-large zero-shot)", "author": "Deepan Wadhwa", "repo": "https://github.com/deepanwadhwa/OpenDecision", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "classifier", "open": "open", "underlying": "MoritzLaurer/ModernBERT-large-zeroshot-v2.0 (~400M) used as a zero-shot NLI decision engine", "rankOfficial": 58, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 21.64, "jevCheese": 66.2, "axes": { "intelligence": 31.8, "calibration": 57.1, "speed": 79.9, "cost": 75.3 }, "usdPer1kDecisions": 0.0066, "costKind": "estimate", "value": 9977.5, "publicAccuracy": 53.2, "sealedAccuracy": 25.6, "gapPp": 27.6, "notRankedBecause": null, "rank": 58, "onPareto": false }, { "slug": "litjev", "name": "LitJev (Qwen3.8-27B)", "author": "Zhengxu Yu", "repo": "https://github.com/zhengxuyu/litjev", "repoHost": "GitHub", "licence": "Apache-2.0 (code); Apache-2.0 base weights", "class": "jev-rebuild", "open": "open", "underlying": "Qwen/Qwen3.8-27B, frozen, read at the output head", "rankOfficial": 59, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 19.51, "jevCheese": 56.49, "axes": { "intelligence": 46.3, "calibration": 76.6, "speed": 66.7, "cost": 33.6 }, "usdPer1kDecisions": 0.163, "costKind": "estimate", "value": 346.5, "publicAccuracy": 86.1, "sealedAccuracy": 30.8, "gapPp": 55.3, "notRankedBecause": null, "rank": 59, "onPareto": false }, { "slug": "openjev-verdict-1.4", "name": "openJev Verdict 1.4", "author": "Hemant (heman10x)", "repo": "https://huggingface.co/heman10x/rlcd-modernbert-151m", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "GLiClass ModernBERT-base fine-tuned decision model, 151M; v1.4 fixed inference engine", "rankOfficial": 60, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 19.0, "jevCheese": 75.06, "axes": { "intelligence": 29.4, "calibration": 72.0, "speed": 78.1, "cost": 82.4 }, "usdPer1kDecisions": 0.0039, "costKind": "estimate", "value": 19380.6, "publicAccuracy": 57.6, "sealedAccuracy": 27.9, "gapPp": 29.7, "notRankedBecause": null, "rank": 60, "onPareto": false }, { "slug": "kev-0.5b", "name": "kev 0.5B", "author": "Jared Palmer", "repo": "https://github.com/jaredpalmer/kev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "Qwen2.5-0.5B + LoRA + learned pointer head; jaredpalmer/kev-0.5b", "rankOfficial": 61, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 18.88, "jevCheese": 62.89, "axes": { "intelligence": 30.5, "calibration": 49.7, "speed": 77.0, "cost": 76.1 }, "usdPer1kDecisions": 0.0063, "costKind": "estimate", "value": 10033.7, "publicAccuracy": 49.4, "sealedAccuracy": 27.3, "gapPp": 22.1, "notRankedBecause": null, "rank": 61, "onPareto": false }, { "slug": "nimble-9b", "name": "Bespoke Nimble 9B (Bespoke Labs)", "author": "Bespoke Labs", "repo": "https://github.com/bespokelabsai/nimble", "repoHost": "GitHub", "licence": "Apache-2.0 (weights); repository without a licence file as of 19 Sep", "class": "jev-rebuild", "open": "open", "underlying": "bespokelabs/Bespoke-Nimble-9B (LoRA, adapter unchanged since 93ec5d6), merged into Qwen/Qwen3.5-9B@c202236 with the author's PEFT safe-merge", "rankOfficial": 62, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 18.66, "jevCheese": 51.36, "axes": { "intelligence": 46.3, "calibration": 56.4, "speed": 78.7, "cost": 33.4 }, "usdPer1kDecisions": 0.1658, "costKind": "estimate", "value": 309.7, "publicAccuracy": 79.7, "sealedAccuracy": 28.9, "gapPp": 50.8, "notRankedBecause": null, "rank": 62, "onPareto": false }, { "slug": "gpt-5.6-luna", "name": "GPT-5.6 Luna (low reasoning effort)", "author": "OpenAI", "repo": null, "repoHost": null, "licence": "proprietary API", "class": "llm-baseline", "open": "closed", "underlying": "closed", "rankOfficial": 63, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 18.51, "jevCheese": 82.46, "axes": { "intelligence": 93.1, "calibration": 87.4, "speed": 77.5, "cost": 28.5 }, "usdPer1kDecisions": 0.2419, "costKind": "measured", "value": 340.9, "publicAccuracy": 97.4, "sealedAccuracy": 89.0, "gapPp": 8.4, "notRankedBecause": null, "rank": 63, "onPareto": false }, { "slug": "openjev-verdict", "name": "openJev Verdict (heman10x, ModernBERT-base 151M)", "author": "Hemant (heman10x)", "repo": "https://github.com/Heman10x-NGU/openJev-verdict-2.0", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "GLiClass ModernBERT-base (knowledgator/gliclass-modern-base-v2.0) fine-tuned, 151M", "rankOfficial": 64, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 18.09, "jevCheese": 61.84, "axes": { "intelligence": 30.0, "calibration": 47.0, "speed": 76.7, "cost": 83.1 }, "usdPer1kDecisions": 0.0037, "costKind": "estimate", "value": 16843.4, "publicAccuracy": 55.4, "sealedAccuracy": 24.7, "gapPp": 30.7, "notRankedBecause": null, "rank": 64, "onPareto": false }, { "slug": "raw-qwen3-1.7b", "name": "Raw Qwen3 1.7B direct logits", "author": "Alibaba Qwen / neutral reproduction", "repo": "https://huggingface.co/Qwen/Qwen3-1.7B", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "raw-logit-control", "open": "open", "underlying": "Qwen3-1.7B BF16", "rankOfficial": 65, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 18.06, "jevCheese": 49.06, "axes": { "intelligence": 33.2, "calibration": 24.3, "speed": 89.7, "cost": 64.9 }, "usdPer1kDecisions": 0.0148, "costKind": "estimate", "value": 3315.9, "publicAccuracy": 54.1, "sealedAccuracy": 26.0, "gapPp": 28.1, "notRankedBecause": null, "rank": 65, "onPareto": false }, { "slug": "reflex-27b", "name": "reflex-27b (Qwen3.8-27B)", "author": "kshetrajna12", "repo": "https://github.com/kshetrajna12/reflex", "repoHost": "GitHub", "licence": "MIT code; Apache-2.0 Qwen weights", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.8-27B, frozen, direct-logit readout averaged across two option orders", "rankOfficial": 66, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 17.84, "jevCheese": 56.93, "axes": { "intelligence": 46.4, "calibration": 77.2, "speed": 67.5, "cost": 32.3 }, "usdPer1kDecisions": 0.1811, "costKind": "estimate", "value": 314.3, "publicAccuracy": 87.0, "sealedAccuracy": 29.5, "gapPp": 57.5, "notRankedBecause": null, "rank": 66, "onPareto": false }, { "slug": "jevact", "name": "JevAct (einptein, jev1-2b-v2)", "author": "einptein", "repo": "https://github.com/fstandhartinger/jevbench/issues/66", "repoHost": "GitHub", "licence": "not stated (the author published only an endpoint and the evaluation client)", "class": "jev-rebuild", "open": "closed", "underlying": "the author's own checkpoint model/jev1-2b-v2/checkpoint-45000-67.53 over a Qwen3.5-2B base, served by the author on his own machine; one request per decision returns a native probability for every offered option", "rankOfficial": 67, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 16.93, "jevCheese": 59.68, "axes": { "intelligence": 29.3, "calibration": 55.1, "speed": 76.5, "cost": 64.3 }, "usdPer1kDecisions": 0.0155, "costKind": "estimate", "value": 3852.4, "publicAccuracy": 61.9, "sealedAccuracy": 23.4, "gapPp": 38.5, "notRankedBecause": null, "rank": 67, "onPareto": false }, { "slug": "djev-thinking", "name": "djev (thinking)", "author": "David Villalon / Maisa", "repo": "https://github.com/Davipar/djev-dev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "google/diffusiongemma-26b-a4b-it, BF16; full generation with thinking enabled", "rankOfficial": 68, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 15.2, "jevCheese": 73.39, "axes": { "intelligence": 71.6, "calibration": 87.8, "speed": 75.2, "cost": 26.9 }, "usdPer1kDecisions": 0.2743, "costKind": "estimate", "value": 267.5, "publicAccuracy": 87.4, "sealedAccuracy": 60.1, "gapPp": 27.4, "notRankedBecause": null, "rank": 68, "onPareto": false }, { "slug": "gliner2-large", "name": "GLiNER2 large (Fastino)", "author": "Fastino", "repo": "https://huggingface.co/fastino/gliner2-large-v1", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "classifier", "open": "open", "underlying": "GLiNER2 large schema-conditioned extractor (DeBERTa-v3-large encoder)", "rankOfficial": 69, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 15.14, "jevCheese": 46.38, "axes": { "intelligence": 31.1, "calibration": 24.8, "speed": 61.7, "cost": 73.3 }, "usdPer1kDecisions": 0.0077, "costKind": "estimate", "value": 5988.2, "publicAccuracy": 56.7, "sealedAccuracy": 28.6, "gapPp": 28.1, "notRankedBecause": null, "rank": 69, "onPareto": false }, { "slug": "openjev-thinking", "name": "OpenJev (thinking, BF16)", "author": "razorback16", "repo": "https://github.com/razorback16/openjev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "google/diffusiongemma-26b-a4b-it, BF16; OpenJev think=512", "rankOfficial": 70, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 14.83, "jevCheese": 58.09, "axes": { "intelligence": 58.1, "calibration": 58.1, "speed": 76.1, "cost": 27.8 }, "usdPer1kDecisions": 0.2546, "costKind": "estimate", "value": 228.1, "publicAccuracy": 88.7, "sealedAccuracy": 42.2, "gapPp": 46.5, "notRankedBecause": null, "rank": 70, "onPareto": false }, { "slug": "mghafiri-qwen3.5-0.8b-decision-model", "name": "Qwen3.5-0.8B Decision Model (Mourad Ghafiri)", "author": "Mourad Ghafiri", "repo": "https://huggingface.co/mghafiri/qwen3.5-0.8B-decision-model", "repoHost": "Hugging Face", "licence": "MIT code and training data; Apache-2.0 model weights", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-0.8B-Base fine-tuned as a JevLite decision model with per-question calibration", "rankOfficial": 71, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 14.54, "jevCheese": 58.69, "axes": { "intelligence": 28.1, "calibration": 68.2, "speed": 49.2, "cost": 75.7 }, "usdPer1kDecisions": 0.0065, "costKind": "estimate", "value": 9079.0, "publicAccuracy": 59.3, "sealedAccuracy": 34.7, "gapPp": 24.6, "notRankedBecause": null, "rank": 71, "onPareto": false }, { "slug": "gemini-3.1-flash-lite", "name": "Gemini 3.1 Flash-Lite", "author": "Google", "repo": null, "repoHost": null, "licence": "proprietary API", "class": "llm-baseline", "open": "closed", "underlying": "closed", "rankOfficial": 72, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 14.26, "jevCheese": 56.88, "axes": { "intelligence": 54.5, "calibration": 59.3, "speed": 81.8, "cost": 27.4 }, "usdPer1kDecisions": 0.2638, "costKind": "measured", "value": 215.6, "publicAccuracy": 87.0, "sealedAccuracy": 38.6, "gapPp": 48.4, "notRankedBecause": null, "rank": 72, "onPareto": false }, { "slug": "open-jev-deberta-v3-large", "name": "open-jev-deberta-v3-large (local CPU)", "author": "Kotoba Labs", "repo": "https://github.com/kotoba-lang/typed-decisions", "repoHost": "GitHub", "licence": "Apache-2.0 (model card); DeBERTa-v3 keeps its own terms", "class": "jev-rebuild", "open": "open", "underlying": "microsoft/deberta-v3-large", "rankOfficial": 73, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 12.65, "jevCheese": 66.29, "axes": { "intelligence": 25.6, "calibration": 66.6, "speed": 66.0, "cost": 74.0 }, "usdPer1kDecisions": 0.0073, "costKind": "estimate", "value": 9030.8, "publicAccuracy": 52.4, "sealedAccuracy": 29.5, "gapPp": 22.8, "notRankedBecause": null, "rank": 73, "onPareto": false }, { "slug": "smalljev", "name": "smalljev semantic-v9", "author": "Aditya (isHeSatoshi)", "repo": "https://github.com/isHeSatoshi/smalljev", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "jev-rebuild", "open": "open", "underlying": "MiniCPM5-2B-Base, 2.5B dense, with LoRA and native semantic decision heads", "rankOfficial": 74, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 12.31, "jevCheese": 58.52, "axes": { "intelligence": 25.7, "calibration": 59.2, "speed": 79.8, "cost": 57.9 }, "usdPer1kDecisions": 0.0254, "costKind": "estimate", "value": 2307.2, "publicAccuracy": 60.6, "sealedAccuracy": 26.9, "gapPp": 33.7, "notRankedBecause": null, "rank": 74, "onPareto": false }, { "slug": "gliner2", "name": "GLiNER2 (Fastino, gliner2.5-base)", "author": "Fastino", "repo": "https://github.com/fastino-ai/GLiNER2", "repoHost": "GitHub", "licence": "Apache-2.0", "class": "classifier", "open": "open", "underlying": "DeBERTa-v3-base schema-conditioned extractor (GLiNER2.5 boundary architecture), 194M", "rankOfficial": 75, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 11.78, "jevCheese": 49.62, "axes": { "intelligence": 27.4, "calibration": 25.2, "speed": 71.8, "cost": 83.1 }, "usdPer1kDecisions": 0.0037, "costKind": "estimate", "value": 13516.7, "publicAccuracy": 58.0, "sealedAccuracy": 29.2, "gapPp": 28.8, "notRankedBecause": null, "rank": 75, "onPareto": false }, { "slug": "instinct", "name": "Instinct (ZooWork, Qwen3.8-27B)", "author": "rayrain-srp (ZooWork)", "repo": "https://github.com/fstandhartinger/jevbench/issues/69", "repoHost": "GitHub", "licence": "proprietary service over Apache-2.0 base weights (Qwen3.8-27B)", "class": "jev-rebuild", "open": "closed", "underlying": "Qwen3.8-27B (Apache-2.0 base weights) with no additional fine-tuning; their own serving stack reads candidate-token scores in one forward pass per question and normalizes them into a decision probability, with no autoregressive decoding. The serving implementation is not public.", "rankOfficial": 76, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 11.45, "jevCheese": 64.91, "axes": { "intelligence": 51.1, "calibration": 78.7, "speed": 83.9, "cost": 24.6 }, "usdPer1kDecisions": 0.3253, "costKind": "estimate", "value": 199.5, "publicAccuracy": 86.6, "sealedAccuracy": 35.1, "gapPp": 51.5, "notRankedBecause": null, "rank": 76, "onPareto": false }, { "slug": "open-jev-zefan-9b", "name": "Open-Jev 9B (Zefan Cai)", "author": "Zefan Cai (@Zefan_Cai)", "repo": "https://github.com/Zefan-Cai/Open-Jev", "repoHost": "GitHub", "licence": "MIT (loader); Apache-2.0 (adapter and pinned Qwen base); CC0-1.0 public training projection", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-9B plus rank-8 LoRA and trained scalar decision head", "rankOfficial": 77, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 11.2, "jevCheese": 52.98, "axes": { "intelligence": 44.2, "calibration": 61.8, "speed": 72.0, "cost": 28.1 }, "usdPer1kDecisions": 0.2488, "costKind": "estimate", "value": 212.9, "publicAccuracy": 77.5, "sealedAccuracy": 29.9, "gapPp": 47.6, "notRankedBecause": null, "rank": 77, "onPareto": false }, { "slug": "open-jev-zefan-2b", "name": "Open-Jev 2B (Zefan Cai)", "author": "Zefan Cai (@Zefan_Cai)", "repo": "https://github.com/Zefan-Cai/Open-Jev", "repoHost": "GitHub", "licence": "MIT (loader); Apache-2.0 (adapter and pinned Qwen base); CC0-1.0 public training projection", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-2B plus rank-8 LoRA and trained scalar decision head", "rankOfficial": 78, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 9.98, "jevCheese": 48.8, "axes": { "intelligence": 42.3, "calibration": 55.3, "speed": 73.5, "cost": 28.1 }, "usdPer1kDecisions": 0.2488, "costKind": "estimate", "value": 196.1, "publicAccuracy": 64.5, "sealedAccuracy": 26.3, "gapPp": 38.2, "notRankedBecause": null, "rank": 78, "onPareto": false }, { "slug": "gliner2.5-multi", "name": "GLiNER2.5 multi (Fastino, 287M)", "author": "Fastino", "repo": "https://huggingface.co/fastino/gliner2.5-multi-v1", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "classifier", "open": "open", "underlying": "GLiNER2.5 multilingual schema-conditioned extractor, 287M", "rankOfficial": 79, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 9.76, "jevCheese": 62.49, "axes": { "intelligence": 23.1, "calibration": 57.2, "speed": 67.8, "cost": 82.4 }, "usdPer1kDecisions": 0.0039, "costKind": "estimate", "value": 16134.7, "publicAccuracy": 48.9, "sealedAccuracy": 32.8, "gapPp": 16.1, "notRankedBecause": null, "rank": 79, "onPareto": false }, { "slug": "clm-8b", "name": "CLM-8B (Contrastive-LM, clm-latest)", "author": "Contrastive-LM (Kwok, Kang, Suresh, Saad-Falcon, Pavone, Ré, Mirhoseini)", "repo": "https://github.com/Contrastive-LM/CLM", "repoHost": "GitHub", "licence": "Apache-2.0 (code and CLM-v0.1-8B head); Qwen3-8B encoder Apache-2.0", "class": "system-one-open", "open": "open", "underlying": null, "rankOfficial": 80, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 8.57, "jevCheese": 59.12, "axes": { "intelligence": 22.4, "calibration": 39.8, "speed": 93.6, "cost": 78.4 }, "usdPer1kDecisions": 0.0052, "costKind": "estimate", "value": 11295.1, "publicAccuracy": 40.7, "sealedAccuracy": 24.0, "gapPp": 16.7, "notRankedBecause": null, "rank": 80, "onPareto": false }, { "slug": "simplejev-qwen3.5-0.8b", "name": "SimpleJev (Qwen3.5-0.8B, CPU)", "author": "sabeel111 / Featherless AI", "repo": "https://github.com/featherless-ai/simple-jev", "repoHost": "GitHub", "licence": "Apache-2.0 server; Apache-2.0 Qwen3.5-0.8B checkpoint", "class": "jev-rebuild", "open": "open", "underlying": "Qwen3.5-0.8B through SimpleJev native assistant-prefill option-logit scorer", "rankOfficial": 81, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 7.46, "jevCheese": 53.26, "axes": { "intelligence": 21.5, "calibration": 49.1, "speed": 57.5, "cost": 68.3 }, "usdPer1kDecisions": 0.0114, "costKind": "estimate", "value": 4657.9, "publicAccuracy": 54.5, "sealedAccuracy": 34.7, "gapPp": 19.8, "notRankedBecause": null, "rank": 81, "onPareto": false }, { "slug": "gliner2.5-small", "name": "GLiNER2.5 small (Fastino, 74M)", "author": "Fastino", "repo": "https://huggingface.co/fastino/gliner2.5-small-v1", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "classifier", "open": "open", "underlying": "GLiNER2.5 small schema-conditioned extractor, 74M", "rankOfficial": 82, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 7.19, "jevCheese": 64.27, "axes": { "intelligence": 20.5, "calibration": 50.7, "speed": 77.8, "cost": 82.4 }, "usdPer1kDecisions": 0.0039, "costKind": "estimate", "value": 16594.9, "publicAccuracy": 45.9, "sealedAccuracy": 28.6, "gapPp": 17.3, "notRankedBecause": null, "rank": 82, "onPareto": false }, { "slug": "raw-qwen3-0.6b", "name": "Raw Qwen3 0.6B direct logits", "author": "Alibaba Qwen / neutral reproduction", "repo": "https://huggingface.co/Qwen/Qwen3-0.6B", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "raw-logit-control", "open": "open", "underlying": "Qwen3-0.6B BF16", "rankOfficial": 83, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 7.14, "jevCheese": 48.37, "axes": { "intelligence": 22.8, "calibration": 20.7, "speed": 89.9, "cost": 73.9 }, "usdPer1kDecisions": 0.0074, "costKind": "estimate", "value": 6538.7, "publicAccuracy": 48.1, "sealedAccuracy": 25.6, "gapPp": 22.4, "notRankedBecause": null, "rank": 83, "onPareto": false }, { "slug": "verdict-small", "name": "verdict-small (Manavarya09, multilingual-e5-small 118M)", "author": "Manavarya09 (Manav)", "repo": "https://github.com/Manavarya09/verdict", "repoHost": "GitHub", "licence": "Apache-2.0 (verdictml code and the Manav2op/verdict-small checkpoint over intfloat/multilingual-e5-small)", "class": "jev-rebuild", "open": "open", "underlying": "intfloat/multilingual-e5-small (118M multilingual bi-encoder) fine-tuned on a typed-decision mix; each option is scored against the rendered state by cosine similarity, with temperature scaling and a conformal abstain set on top. One encoder pass per option, no tokens generated.", "rankOfficial": 84, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 5.69, "jevCheese": 75.95, "axes": { "intelligence": 18.1, "calibration": 66.9, "speed": 85.0, "cost": 96.6 }, "usdPer1kDecisions": 0.0013, "costKind": "estimate", "value": 58675.4, "publicAccuracy": 53.7, "sealedAccuracy": 28.9, "gapPp": 24.8, "notRankedBecause": null, "rank": 84, "onPareto": true }, { "slug": "deepseek-flash", "name": "DeepSeek V4.1 Flash (thinking default)", "author": "DeepSeek", "repo": "https://huggingface.co/deepseek-ai", "repoHost": "Hugging Face", "licence": "open weights, proprietary API route", "class": "llm-baseline", "open": "open-weights", "underlying": "DeepSeek-V4.1-Flash", "rankOfficial": 85, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": true, "jevbenchScore": 4.77, "jevCheese": 82.8, "axes": { "intelligence": 94.0, "calibration": 95.5, "speed": 71.6, "cost": 16.8 }, "usdPer1kDecisions": 0.5937, "costKind": "measured", "value": 139.5, "publicAccuracy": 97.8, "sealedAccuracy": 94.8, "gapPp": 3.0, "notRankedBecause": null, "rank": 85, "onPareto": false }, { "slug": "mirror", "name": "Mirror", "author": "Bluusun", "repo": "https://github.com/Bluusun/Decision-API", "repoHost": "GitHub", "licence": "Apache-2.0 wrapper and Mirror release; upstream DeBERTa/model assets retain their own terms", "class": "jev-rebuild", "open": "open", "underlying": "Mirror DeBERTa-v3-large 436M classification-span scorer", "rankOfficial": 86, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 2.11, "jevCheese": 48.38, "axes": { "intelligence": 13.6, "calibration": 26.0, "speed": 70.8, "cost": 73.3 }, "usdPer1kDecisions": 0.0077, "costKind": "estimate", "value": 6245.7, "publicAccuracy": 41.1, "sealedAccuracy": 9.1, "gapPp": 32.0, "notRankedBecause": null, "rank": 86, "onPareto": false }, { "slug": "mxbai-rerank-base-v2", "name": "Mixedbread mxbai-rerank-base-v2", "author": "Mixedbread", "repo": "https://huggingface.co/mixedbread-ai/mxbai-rerank-base-v2", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "reranker", "open": "open", "underlying": "494M cross-encoder", "rankOfficial": 87, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 0.4, "jevCheese": 76.02, "axes": { "intelligence": 6.8, "calibration": 84.1, "speed": 87.5, "cost": 67.9 }, "usdPer1kDecisions": 0.0117, "costKind": "measured", "value": 6485.3, "publicAccuracy": 37.2, "sealedAccuracy": 34.4, "gapPp": 2.8, "notRankedBecause": null, "rank": 87, "onPareto": false }, { "slug": "bge-reranker-v2-m3", "name": "BAAI bge-reranker-v2-m3", "author": "BAAI", "repo": "https://huggingface.co/BAAI/bge-reranker-v2-m3", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "reranker", "open": "open", "underlying": "568M multilingual cross-encoder", "rankOfficial": 88, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 0.17, "jevCheese": 78.8, "axes": { "intelligence": 5.0, "calibration": 84.2, "speed": 89.5, "cost": 73.4 }, "usdPer1kDecisions": 0.0077, "costKind": "measured", "value": 10208.5, "publicAccuracy": 39.4, "sealedAccuracy": 27.9, "gapPp": 11.5, "notRankedBecause": null, "rank": 88, "onPareto": false }, { "slug": "gte-reranker-modernbert-base", "name": "Alibaba GTE Reranker ModernBERT-base", "author": "Alibaba-NLP", "repo": "https://huggingface.co/Alibaba-NLP/gte-reranker-modernbert-base", "repoHost": "Hugging Face", "licence": "Apache-2.0", "class": "reranker", "open": "open", "underlying": "149M ModernBERT cross-encoder", "rankOfficial": 89, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 0.15, "jevCheese": 74.26, "axes": { "intelligence": 4.8, "calibration": 78.9, "speed": 90.6, "cost": 69.6 }, "usdPer1kDecisions": 0.0103, "costKind": "measured", "value": 7223.9, "publicAccuracy": 33.8, "sealedAccuracy": 33.4, "gapPp": 0.3, "notRankedBecause": null, "rank": 89, "onPareto": false }, { "slug": "certo", "name": "Certo v1 (AltSlate Labs)", "author": "AltSlate Labs", "repo": "https://huggingface.co/altslate/certo-decision-model", "repoHost": "Hugging Face", "licence": "MIT", "class": "jev-rebuild", "open": "open", "underlying": "ModernBERT-large with a per-option query/scoring head, ~400M parameters", "rankOfficial": 90, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 0.0, "jevCheese": 88.48, "axes": { "intelligence": 0.1, "calibration": 83.0, "speed": 94.0, "cost": 100.0 }, "usdPer1kDecisions": 0.001, "costKind": "estimate", "value": 91645.0, "publicAccuracy": 31.6, "sealedAccuracy": 29.5, "gapPp": 2.1, "notRankedBecause": null, "rank": 90, "onPareto": true }, { "slug": "open-jev-json-canvas-joshuasp", "name": "Open Jev JSON Canvas (JoshuaSP)", "author": "JoshuaSP", "repo": "https://github.com/JoshuaSP/open-jev", "repoHost": "GitHub", "licence": "MIT code; Apache-2.0 DiffusionGemma weights", "class": "jev-rebuild", "open": "open", "underlying": "google/diffusiongemma-26B-A4B-it BF16, one-step JSON canvas", "rankOfficial": 91, "ranked": true, "listing": "ranked", "partial": false, "apiFlag": false, "jevbenchScore": 0.0, "jevCheese": 46.89, "axes": { "intelligence": 48.2, "calibration": 0.0, "speed": 84.1, "cost": 45.6 }, "usdPer1kDecisions": 0.065, "costKind": "estimate", "value": 721.6, "publicAccuracy": 84.4, "sealedAccuracy": 31.2, "gapPp": 53.2, "notRankedBecause": null, "rank": 91, "onPareto": false }, { "slug": "classifier-dev-fast", "name": "classifier.dev (fast tier)", "author": "mrmps (@michael_chomsky)", "repo": "https://classifier.dev", "repoHost": "classifier.dev", "licence": "MIT (code); hosted service", "class": "jev-service", "open": "closed", "underlying": "Jev (TypeSafe) behind classifier.dev's zero-shot classification API", "rankOfficial": null, "ranked": false, "listing": "honorable_mention", "partial": false, "apiFlag": true, "jevbenchScore": 70.82, "jevCheese": 78.37, "axes": { "intelligence": 51.6, "calibration": 72.4, "speed": 87.6, "cost": 84.3 }, "usdPer1kDecisions": 0.0033, "costKind": "estimate", "value": 23510.9, "publicAccuracy": 85.3, "sealedAccuracy": 34.4, "gapPp": 50.9, "notRankedBecause": "runs on Jev (TypeSafe) — listed, not ranked", "rank": null, "onPareto": false }, { "slug": "qwen3.8-27b", "name": "Qwen3.8 27B (Chutes TEE)", "author": "Qwen / Chutes", "repo": null, "repoHost": null, "licence": "open weights", "class": "llm-baseline", "open": "open-weights", "underlying": "Qwen3.8-27B", "rankOfficial": null, "ranked": false, "listing": "partial", "partial": true, "apiFlag": true, "jevbenchScore": 0.0, "jevCheese": 50.84, "axes": { "intelligence": 40.4, "calibration": 93.6, "speed": 61.3, "cost": 0.0 }, "usdPer1kDecisions": 2.6691, "costKind": "estimate", "value": 19.0, "publicAccuracy": 71.9, "sealedAccuracy": 21.8, "gapPp": 50.1, "notRankedBecause": null, "rank": null, "onPareto": false }, { "slug": "needle-3-tools", "name": "Needle 3, options as tools (post-hoc adapter mode)", "author": "Cactus Compute", "repo": "https://github.com/cactus-compute/needle", "repoHost": "GitHub", "licence": "Apache-2.0 (model and package)", "class": "small-tool-model", "open": "open", "underlying": "Needle 3 (121M parameters, 2-bit)", "rankOfficial": null, "ranked": false, "listing": "partial", "partial": true, "apiFlag": false, "jevbenchScore": null, "jevCheese": null, "axes": { "intelligence": null, "calibration": null, "speed": null, "cost": null }, "usdPer1kDecisions": 0.0144, "costKind": "estimate", "value": null, "publicAccuracy": 22.1, "sealedAccuracy": null, "gapPp": null, "notRankedBecause": "No completed sealed v1.4 run; retained as an unranked historical row", "rank": null, "onPareto": false }, { "slug": "needle-3", "name": "Needle 3 (Cactus, 2-bit, local CPU)", "author": "Cactus Compute", "repo": "https://github.com/cactus-compute/needle", "repoHost": "GitHub", "licence": "Apache-2.0 (model and package)", "class": "small-tool-model", "open": "open", "underlying": "Needle 3 (121M parameters, 2-bit)", "rankOfficial": null, "ranked": false, "listing": "partial", "partial": true, "apiFlag": false, "jevbenchScore": null, "jevCheese": null, "axes": { "intelligence": null, "calibration": null, "speed": null, "cost": null }, "usdPer1kDecisions": 0.0238, "costKind": "estimate", "value": null, "publicAccuracy": 22.5, "sealedAccuracy": null, "gapPp": null, "notRankedBecause": "No completed sealed v1.4 run; retained as an unranked historical row", "rank": null, "onPareto": false } ] }