Jev Decision Index Leaderboard
decision160 models · Updated October 2026
On the Jev Decision Index benchmark, vLLM-SR d3 (27B) ranks #1 with a score of 66.3. The full ranking, with cost per million tokens, is below.
| # | Model | Score | ECE | p50 latency | Input / 1M |
|---|---|---|---|---|---|
| 1 | vLLM-SR d3 (27B)vLLM Semantic Router | 66.3 | 0.043 | 97 ms | — |
| 2 | Decider 31B (NVFP4)Mapika | 64.6 | 0.024 | 72 ms | — |
| 3 | Perplexity Decider v1.1 (27B)Perplexity | 62.8 | 0.071 | 104 ms | $0.02 |
| 4 | Darwin-27B-ZTCFINAL_Bench | 62.5 | 0.045 | 104 ms | — |
| 5 | AnyJev L0 · Gemma-4-31Bmorriszjm | 60.2 | 0.240 | 98 ms | — |
| 6 | Fastino GLiDEFastino | 60.2 | 0.040 | 98 ms | $0.15 |
| 7 | Sieve 27BSthānika AI | 60.2 | 0.065 | 101 ms | — |
| 8 | JevTypeSafe AI | 60.1 | 0.074 | 524 ms | $0.04 |
| 9 | Torchcast Decision 27BTorchcast AI | 59.9 | 0.024 | 100 ms | — |
| 10 | Shisa DE-2 (repeat)Shisa.AI | 59.8 | 0.207 | 57 ms | — |
| 11 | Matilda Jev v1.5Maincode | 59.2 | 0.077 | 89 ms | — |
| 12 | deck31b (Gemma 4 31B)krishna-gogineni-765 | 59.0 | 0.204 | 219 ms | — |
| 13 | Kev 27BJared Palmer | 58.8 | 0.022 | 135 ms | — |
| 14 | Quyet-1.0-Largechinhnc | 58.7 | 0.078 | 105 ms | — |
| 15 | RSI-Jev v6.1 VL 27B (effort auto)shgao | 58.7 | 0.077 | 76 ms | — |
| 16 | decisio 0.6.0 (Gemma 4 31B)aminry | 58.5 | 0.142 | 54 ms | — |
| 17 | Qwen Decision 27B noncausal LoRAjsaurabh | 58.5 | 0.051 | 127 ms | — |
| 18 | garas (Gemma 4 31B)plantwaterer | 58.3 | 0.049 | 117 ms | — |
| 19 | Wald-26B-A4BORG2 AI | 58.3 | 0.043 | 90 ms | — |
| 20 | linne-fern 1.0 (27B)MISA-AICORE | 57.8 | 0.079 | 108 ms | — |
| 21 | Blink v0.3 26B-A4B (NVFP4)Pixilab.ai | 57.8 | 0.040 | 25 ms | — |
| 22 | Decider chat · Gemma-4-31BMapika | 57.6 | 0.047 | 109 ms | — |
| 23 | SynACK Decide 26B-A4B v7SkyPanther | 57.5 | 0.069 | 129 ms | — |
| 24 | Surogate Rune 26B-A4B v3Surogate | 57.4 | 0.120 | 121 ms | — |
| 25 | Xor 26B-A4B v1.1Juspay Technologies | 57.1 | 0.058 | 110 ms | — |
| 26 | AJev lora1 (Gemma 4 26B-A4B)andyzhang232 | 56.4 | 0.048 | 89 ms | — |
| 27 | GEV-26B-DecideAutoTrust AI Lab | 56.2 | 0.072 | 129 ms | — |
| 28 | Nace.AI Drex v2.0 (35B-A3B)Nace.AI | 55.9 | 0.018 | 276 ms | $0.06 |
| 29 | frontier-infra Jebadiah 27B v2.1Frontier Infra | 55.9 | 0.061 | 110 ms | — |
| 30 | vLLM-SR Decision 2.0 Vega 27BvLLM Semantic Router | 55.9 | 0.086 | 120 ms | — |
| 31 | vLLM-SR d3-flash (9B)vLLM Semantic Router | 55.8 | 0.056 | 54 ms | — |
| 32 | JEV-27BAutoTrust AI Lab | 55.6 | 0.076 | 102 ms | — |
| 33 | JADEtheunnecessarythings | 55.6 | 0.238 | 133 ms | — |
| 34 | Eikos-27Bcaiovicentino1 | 55.3 | 0.057 | 130 ms | — |
| 35 | reflex Qwen3.8-27B-FP8 (pmscr)kshetrajna12 | 54.7 | 0.045 | 149 ms | — |
| 36 | Bespoke Nimble 9B v3Bespoke Labs | 54.7 | 0.123 | 74 ms | — |
| 37 | OneJev1.5-9BOmniJev | 53.3 | 0.029 | 111 ms | — |
| 38 | Cloudflare clefCloudflare | 53.1 | 0.040 | 102 ms | — |
| 39 | vLLM-SR d3-mini (4B)vLLM Semantic Router | 52.6 | 0.056 | 50 ms | — |
| 40 | simple-jev · Qwen3.8-27B (featherless)featherless-ai | 51.8 | 0.113 | 373 ms | — |
| 41 | JPT-35B-A3Bkirp | 51.4 | 0.049 | 250 ms | — |
| 42 | Decider chat · Qwen3.6-27BMapika | 51.3 | 0.021 | 84 ms | — |
| 43 | Xor 1.2Juspay Technologies | 51.1 | 0.024 | 150 ms | — |
| 44 | Sieve 9B v2Sthānika AI | 50.9 | 0.073 | 51 ms | — |
| 45 | AJev lora7 (Gemma 4 12B)andyzhang232 | 50.7 | 0.082 | 69 ms | — |
| 46 | Nace.AI Drex v1.5Nace.AI | 49.7 | 0.090 | 54 ms | $0.05 |
| 47 | Winnow-12BEldanRing | 49.3 | 0.168 | 73 ms | — |
| 48 | CoCo-Decision-4B 2.3.0Corners | 48.8 | 0.116 | 63 ms | — |
| 49 | Decider 35B-A3BMapika | 48.5 | 0.023 | 101 ms | — |
| 50 | JevK5 9B v0.3.3alibiserikbay | 48.4 | 0.060 | 40 ms | — |
| 51 | Cloudflare clef-omni (30B-A3B)Cloudflare | 48.3 | 0.040 | 191 ms | $0.15 |
| 52 | vLLM-SR Decision 2.0 Lux 9BvLLM Semantic Router | 48.2 | 0.056 | 37 ms | — |
| 53 | decisio 0.4.0 (Qwen3.6-35B-A3B)aminry | 47.7 | 0.055 | 71 ms | — |
| 54 | Cloudflare clef-flashCloudflare | 47.6 | 0.025 | 53 ms | — |
| 55 | decisio 0.4.0 (Gemma 4 12B)aminry | 47.5 | 0.103 | 38 ms | — |
| 56 | JoshuaSP diffusiongemma (open-jev)JoshuaSP | 47.3 | 0.215 | 260 ms | — |
| 57 | Hopper 12B (trained)HopitAI | 47.3 | 0.076 | 290 ms | — |
| 58 | ezjev 4B s2everettjf | 47.0 | 0.063 | 36 ms | — |
| 59 | PurpleMIST-Flash-1.0Olaverse Lab | 46.8 | 0.032 | 53 ms | — |
| 60 | RSI-Jev v6.1 VL 4Bshgao | 46.3 | 0.049 | 58 ms | — |
| 61 | Sieve 4BSthānika AI | 46.1 | 0.031 | 93 ms | — |
| 62 | Jevfirekikoncuo | 45.5 | 0.052 | 90 ms | — |
| 63 | JEV-9BAutoTrust AI Lab | 45.2 | 0.073 | 50 ms | — |
| 64 | Standard One 8B SH (v3)Standard Thinking | 45.0 | 0.080 | 80 ms | — |
| 65 | vLLM-SR Decision 2.0 Nox 4BvLLM Semantic Router | 45.0 | 0.030 | 34 ms | — |
| 66 | frontier-infra Jebadiah 9B v2.1Frontier Infra | 44.9 | 0.058 | 49 ms | — |
| 67 | jiwo 4B v2eljiwo | 44.8 | 0.068 | 50 ms | — |
| 68 | Sieve 9BSthānika AI | 44.6 | 0.043 | 50 ms | — |
| 69 | JPT-9Bkirp | 44.5 | 0.077 | 149 ms | — |
| 70 | Aplomb 1EmpirioLabs AI | 43.8 | 0.075 | 54 ms | — |
| 71 | FlyMy.AI Decision 4B v1.6FlyMy.AI | 43.6 | 0.046 | 27 ms | — |
| 72 | Kev 9B v2Jared Palmer | 43.3 | 0.045 | 60 ms | — |
| 73 | Hopper (G) 1.2HopitAI | 42.6 | 0.093 | 23 ms | — |
| 74 | Jev-Omniakhilaaa3 | 42.4 | 0.161 | 55 ms | — |
| 75 | djevDavipar | 41.8 | 0.212 | 84 ms | — |
| 76 | JPT-4Bkirp | 41.6 | 0.079 | 139 ms | — |
| 77 | JevK5 v0.3alibiserikbay | 41.5 | 0.060 | 26 ms | — |
| 78 | Bespoke Nimble 9B v2Bespoke Labs | 41.0 | 0.024 | 77 ms | — |
| 79 | Solomon v1.1Doccy Health | 40.9 | 0.081 | 224 ms | — |
| 80 | Decider 4BMapika | 40.3 | 0.084 | 13 ms | — |
| 81 | Saracura PT-BR 4BFelhen | 40.2 | 0.179 | 62 ms | — |
| 82 | razorback16 openjev diffusiongemma (NVFP4, vLLM)razorback16 | 40.0 | 0.232 | 38 ms | — |
| 83 | Standard One 3B SH (v2)Standard Thinking | 39.9 | 0.101 | 76 ms | — |
| 84 | Kev 4B r10Jared Palmer | 39.5 | 0.032 | 59 ms | $0.04 |
| 85 | LiquidAI d1-3BLiquid AI | 38.9 | 0.047 | 16 ms | — |
| 86 | vLLM-SR d3-nano (2B)vLLM Semantic Router | 38.9 | 0.112 | 38 ms | — |
| 87 | Winnow-E4BEldanRing | 38.7 | 0.058 | 45 ms | — |
| 88 | Lev1 3.7Btheaviv | 38.6 | 0.058 | 29 ms | — |
| 89 | RSI-Jev v5.0 VL 3Bshgao | 38.4 | 0.048 | 19 ms | — |
| 90 | Intern-Decision-4BIntern Large Models | 38.2 | 0.028 | 44 ms | — |
| 91 | NeoHorse-Jev-4BTokenRhythm | 38.0 | 0.104 | 49 ms | — |
| 92 | Ryotide Qwen v0.3.1csabag | 37.0 | 0.092 | 81 ms | — |
| 93 | Kas 4Bkpiya | 36.7 | 0.098 | 69 ms | — |
| 94 | levInterfaze | 36.5 | 0.064 | 71 ms | — |
| 95 | mmastrac diffusiongemma (vLLM PR 57250)mmastrac | 36.3 | 0.202 | 127 ms | — |
| 96 | Jevstral 8BAmirBraham | 35.9 | 0.047 | 28 ms | — |
| 97 | Intelif Qwen3-4BUserMoonlight | 34.3 | 0.071 | 23 ms | — |
| 98 | Tev1-4B-experimentalTogether | 34.1 | 0.104 | 36 ms | $0.04 |
| 99 | JobeMantisShrimpdev | 33.1 | 0.123 | 54 ms | — |
| 100 | Gevva E4Bdavidburhans | 33.0 | 0.145 | 53 ms | — |
| 101 | vLLM-SR d3-lite (0.8B)vLLM Semantic Router | 32.3 | 0.094 | 37 ms | — |
| 102 | openvonsgenai-craft | 32.3 | 0.371 | 21 ms | — |
| 103 | open-jev (pngwn)pngwn | 32.1 | 0.059 | 112 ms | — |
| 104 | Jev-Style 2B v3chaoliangUNSW | 31.4 | 0.118 | 70 ms | — |
| 105 | Metask-Jev-4BMetask-AI-Lab | 31.0 | 0.103 | 57 ms | — |
| 106 | Sieve 2B v2Sthānika AI | 30.8 | 0.095 | 68 ms | — |
| 107 | Cam1cbuk10001 | 30.8 | 0.063 | 249 ms | — |
| 108 | OneJev1.5-0.8BOmniJev | 29.0 | 0.035 | 83 ms | — |
| 109 | SemIfTheoLeeCJ | 28.8 | 0.099 | 113 ms | — |
| 110 | Gevva E2Bdavidburhans | 28.2 | 0.090 | 43 ms | — |
| 111 | vera-corescar-ai | 28.1 | 0.122 | 100 ms | — |
| 112 | vLLM-SR Decision 2.0 Sol 2BvLLM Semantic Router | 27.1 | 0.037 | 26 ms | — |
| 113 | Decider 2BMapika | 25.8 | 0.077 | 8 ms | — |
| 114 | mini-jevr-ms | 25.2 | 0.346 | 67 ms | — |
| 115 | jiwo 0.8B v2eljiwo | 24.9 | 0.068 | 36 ms | — |
| 116 | this-that 1.2FLock.io | 24.6 | 0.198 | 44 ms | — |
| 117 | EXAONE-4.0-1.2B-JEV v0.3carrtesy | 23.3 | 0.079 | 149 ms | — |
| 118 | Sieve 0.8BSthānika AI | 23.1 | 0.088 | 52 ms | — |
| 119 | Intern-Decision-2BIntern Large Models | 22.6 | 0.061 | 34 ms | — |
| 120 | Crivo-150Mtardellirs | 22.2 | 0.143 | 13 ms | — |
| 121 | Strands Decider 2BStrands Agents | 21.3 | 0.044 | 52 ms | — |
| 122 | Saina Helm 2.0Saina | 20.8 | 0.247 | 39 ms | — |
| 123 | Sieve 2BSthānika AI | 20.5 | 0.088 | 63 ms | — |
| 124 | vLLM-SR d3-edge (0.6B)vLLM Semantic Router | 20.2 | 0.141 | 19 ms | — |
| 125 | JPT-0.8Bkirp | 18.7 | 0.069 | 118 ms | — |
| 126 | Sifr 0.8B v3.1mohamedlotfy50 | 17.8 | 0.120 | 35 ms | — |
| 127 | tasksource Decider nano v4 (150M)tasksource | 17.6 | 0.069 | 14 ms | — |
| 128 | vLLM-SR Decision 2.0 Eos 0.8BvLLM Semantic Router | 17.5 | 0.130 | 25 ms | — |
| 129 | DecisionTune 1.0DecisionTune | 16.7 | 0.147 | 14 ms | — |
| 130 | Bosun v3.1 1.7BHanno Labs | 15.8 | 0.130 | 40 ms | — |
| 131 | Bekko System One v0 400Mhotchpotch | 15.2 | 0.111 | 17 ms | — |
| 132 | Dinah-0Lukitaduarte | 13.9 | 0.104 | 14 ms | — |
| 133 | Kev 0.8B r15Jared Palmer | 13.7 | 0.040 | 45 ms | — |
| 134 | vLLM-SR Decision 2.0 Kai 0.6BvLLM Semantic Router | 13.6 | 0.038 | 19 ms | — |
| 135 | Tev1-0.8B-experimentalTogether | 12.0 | 0.126 | 29 ms | — |
| 136 | Jev-Style 0.8B v3chaoliangUNSW | 11.9 | 0.084 | 36 ms | — |
| 137 | Crivo-32Mtardellirs | 11.9 | 0.087 | 7 ms | — |
| 138 | Kodiak v0.2 1B (accuracy)Cortex Agent LLC | 11.8 | 0.077 | 142 ms | — |
| 139 | Intern-Decision-0.8BIntern Large Models | 11.8 | 0.025 | 34 ms | — |
| 140 | vera-sparkscar-ai | 11.5 | 0.157 | 80 ms | — |
| 141 | GLiNER2.5-DecideFastino | 9.2 | 0.088 | 23 ms | — |
| 142 | LiquidAI d1-omni-600MLiquid AI | 9.1 | 0.128 | 11 ms | — |
| 143 | Bosun v3.1 0.6BHanno Labs | 8.7 | 0.142 | 39 ms | — |
| 144 | Bekko System One v0 68Mhotchpotch | 8.4 | 0.113 | 12 ms | — |
| 145 | Vonwfzyx | 8.2 | 0.216 | 18 ms | — |
| 146 | jefflogan-markewich | 7.7 | 0.097 | 22 ms | — |
| 147 | JevK5 Litealibiserikbay | 6.3 | 0.149 | 19 ms | — |
| 148 | CLM-v0.1-8BContrastive-LM | 6.2 | 0.323 | 47 ms | — |
| 149 | LFM2.5-2.6B-RLCDmonotykamary | 6.1 | 0.255 | 39 ms | — |
| 150 | LAYAConvai Innovations | 4.4 | 0.140 | 6 ms | $0.00 |
| 151 | LavoirMogan AI | 4.4 | 0.149 | 20 ms | — |
| 152 | DM-JEPA 1.1Danger Labs | 4.0 | 0.445 | 63 ms | — |
| 153 | Qwen-2.5-1B-RLCDharshatheg | 3.9 | 0.205 | 44 ms | — |
| 154 | Bekko System One v0 17Mhotchpotch | 2.9 | 0.161 | 7 ms | — |
| 155 | Julia 1Supersonic Labs | 2.6 | 0.420 | 6 ms | — |
| 156 | system-one-gemmaakash-kamat | 1.3 | 0.239 | 32 ms | — |
| 157 | LFM2.5-350M-RLCDnotnotsamuel | 0.0 | 0.568 | 27 ms | — |
| 158 | Nirnaya 0.4Bkshitijthakkar | 0.0 | 0.046 | 14 ms | — |
| 159 | RLCD ModernBERT 151Mheman10x | 0.0 | 0.154 | 11 ms | — |
| 160 | Zevbhubbard | 0.0 | 0.138 | 1 ms | — |
Frequently asked questions
What is the Jev Decision Index benchmark?
Jev Decision Index is a decision used to evaluate and rank large language models. This page lists 160 models by their Jev Decision Index score.
Which model is best on Jev Decision Index?
vLLM-SR d3 (27B) (vLLM Semantic Router) currently leads the Jev Decision Index leaderboard with a score of 66.3.
What is the best value model on Jev Decision Index?
Pricing data is being matched to these models.
Pricing is indicative — confirm with the provider before production use. Updated October 2026.