whatmodel.sh / decision / razorback16 diffgemma (decision readout)

razorback16 diffgemma (decision readout)

+ contribute

razorback16 diffgemma (decision readout) is a source-available decision implementation from vllm-project. Its published size is 25.82B parameters.

Catalogue identity notes

Exact Decision Index tested engine razorback-one-read-3e296f08; reported model razorback16 diffgemma. Type: inference technique; technique: diffusion. Published checkpoint/base reference: nvidia/diffusiongemma-26B-A4B-it-NVFP4. Implementation: https://github.com/vllm-project/vllm/pull/57250. Comparison entry; no current run command, API price or speed inferred from the benchmark.

Model information sources

· Parameter count ↗

Publisher parameter count; includes the served backbone, not just adapter size.

47.9±1.1#222 of 355 decision modelsfrom 54 benchmark results across 2 sources

$ how it compares

The decision models ranked either side of razorback16 diffgemma (decision readout), out of 357 with a score, then more from vllm-project.

#	model	score	where	released
222	InternLM Intern-Decision Qwen3.5-4B Shanghai AI Laboratory (InternLM)	48±1	model files
223	Solar Decide Flash Upstage	48±1	api price unknown
224	razorback16 diffgemma (decision readout) vllm-project	48±1	source code
225	Interfaze lev Qwen3.5-4B LoRA Interfaze	48±1	model files
226	Open-Jev-2B (DecisionBench 0c7aa498) ZefanCai	48±2	model files	2026-09-20
	more from vllm-project
221	mmastrac diffgemma (decision readout) vllm-project	48±1	source code

see all decision models · find the ones that fit your machine

$ what it's good at

tools & automation▁▃▁▁▁▂▃▄▆██▆ #112 of 250 · 7.5% of its score · 6 tests

Beats Senova-4B v2, Tern-35B-A3B, GEV-26B-Decide here, though they rank higher overall. Trails OneJev1.6-0.8B, TokenRhythm NeoHorse-Jev-4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: BFCL (case exact accuracy) submitted	95.7% ✎	▁▂▁▁▁▂▂▄▅▆█▁ beats 142/246	very easy · avg 92%	▇ 2.6%	sharp · noisy
Decision Index 0.4.1: API-Bank (accuracy) submitted	57.3% ✎	▁▁▁▁▁▁▁▂▃▂█▂ beats 66/236	easy · avg 70%	▇ 2.3%	sharp · at its level · noisy
Decision Index 0.4.1: When2Call (accuracy) submitted	51.8% ✎	▁▁▁▂▃▅▅▆█▇█▃ beats 45/246	easy · avg 72%	▇ 1.0%	at its level · noisy
Decision Index 0.4.1: ToolRet (nDCG@10) submitted	46.6% ✎	▁▁▁▁▁·▁▃▁▄█▁ beats 24/246	medium · avg 60%	▇ 0.8%	blunt · at its level
Decision Index 0.4.1: Home appliance simulator (case exact accuracy) submitted	0.0% ✎	█▁▁▃▄▄▆▄▁▂▁▃ beats 47/243	medium · avg 37%	▇ 0.8%	sharp · at its level · noisy
Decision Index 0.4.1: RouterBench (selected quality (quality objective)) submitted	79.9% ✎	no other model yet		—	failed run: Publisher excludes this protocol because its prompt reveals GPT-4 as the best route. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json

retrieval & classification▁▁▁·▁▁▁▁▃▃█▅ #101 of 258 · 5.3% of its score · 7 tests

Beats AutoTrust-JEV-27B v1.1, Decider 31B (NVFP4), Quyet-1.0-Large here, though they rank higher overall. Trails Jev-Style 2B v4, Sieve 2B v3 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: BRIGHT (nDCG@10) submitted	23.2% ✎	▁▁▁▁▁▃▁▂▂▄▇█ beats 26/246	medium · avg 38%	▇ 1.8%	blunt · at its level
Decision Index 0.4.1: Amazon ESCI (macro-F1) submitted	56.3% ✎	▁▁▁▁▁▁▂▂▃▃█▅ beats 171/246	medium · avg 47%	▇ 1.7%	at its level
Decision Index 0.4.1: HoVer (accuracy) submitted	68.4% ✎	▆▆▇▅▄▄▃▇▄█▆▄ beats 94/246	easy · avg 74%	▇ 0.5%	blunt · at its level
Decision Index 0.4.1: PhishNChips phishing decisions (accuracy) submitted	81.3% ✎	█▅▅▄▃▂▁▁▁▁·▁ beats 178/245	easy · avg 70%	▇ 0.5%	at its level · noisy
Decision Index 0.4.1: CLINC150+OOS (macro-F1) submitted	2.0% ✎	▁▃▁▁▁▁▁▂▃▆█▄ beats 4/238	easy · avg 83%	▇ 0.4%	sharp · noisy
Decision Index 0.4.1: BANKING77 (macro-F1) submitted	69.0% ✎	▁·▁▁▁▁▁▁▂█▇█ beats 37/238	easy · avg 79%	▇ 0.3%	noisy
Decision Index 0.4.1: SGD/SGD-X (macro-F1) submitted	34.0% ✎	no other model yet		—	failed run: Publisher reports a builder bug exposing dialogue turns after the classified turn; awaiting corrected rerun. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json

language understanding▂▂▂▃▂▅▆▃█▆▄▃ #169 of 256 · 21% of its score · 9 tests

Beats Hopper 12B (trained), Nari Decision 4B v3.2, Kev 4B r10 here, though they rank higher overall. Trails DecisionTune 1.0, Decider 2B, vLLM-SR d3-lite (0.8B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: HellaSwag (accuracy) submitted	81.1% ✎	▂▄▂▂▄▇▃▇█▅▃▁ beats 88/246	very easy · avg 85%	▇▇▇ 6.3%	sharp
Decision Index 0.4.1: NLI4CT (macro-F1) submitted	74.1% ✎	▁▁▁▂▂▃▃▄▇▇█▄ beats 93/246	easy · avg 74%	▇▇ 4.9%	at its level
Decision Index 0.4.1: ANLI (macro-F1) submitted	56.9% ✎	▁▂▃▅▅█▂····▁ beats 95/246	medium · avg 60%	▇▇ 4.5%	at its level
Decision Index 0.4.1: iSarcasmEval (Sarcasm F1 · track A, English) submitted	56.4% ✎	▁·▁·▁▁▂▄▅█▇▄ beats 168/246	medium · avg 44%	▇ 1.8%	at its level
Decision Index 0.4.1: ContractNLI (macro-F1) submitted	54.1% ✎	▁▁▁▁▁▁▁▃▂▇▇█ beats 37/246	easy · avg 71%	▇ 1.0%	at its level · noisy
Decision Index 0.4.1: VAST (macro-F1) submitted	59.2% ✎	▁▁▁▂▅▄▄▃▄▆█▄ beats 104/246	medium · avg 62%	▇ 0.9%	at its level
Decision Index 0.4.1: ACOS (per-review F1) submitted	14.8% ✎	▁··▂▃▃▅▇█▇▃▃ beats 79/246	hard · avg 22%	▇ 0.7%	noisy
Decision Index 0.4.1: RAGTruth (F1 on hallucinated class) submitted	70.8% ✎	▁▁▁▁▁▁▁▁▁▃▆█ beats 100/246	easy · avg 66%	▇ 0.7%	at its level · noisy

1 more test

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: FinEntity (macro-F1) submitted	85.0% ✎	▁···▁▁▁▁▃▆█▃ beats 99/246	easy · avg 84%	▇ 0.7%	noisy

arts & human taste▁▇▄▃▅▅█▅▄▂▁▁ #120 of 247 · 7.8% of its score · 6 tests

Beats Bobcat 1.3 (31B), JEV-35B, AJev lora1 (Gemma 4 26B-A4B) here, though they rank higher overall. Trails vLLM-SR vela3-fast-vision (Vela 3.0, 0.25B), Crivo-400M, Amber 0.4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: New Yorker (accuracy) submitted	57.6% ✎	▁▁▂▃▃▃▇▅█▅▃▁ beats 85/246	medium · avg 61%	▇ 3.4%	at its level
Decision Index 0.4.1: Humicroedit (accuracy) submitted	60.2% ✎	▁·▁▃▂▂▅█▄▂▁▁ beats 99/246	medium · avg 60%	▇ 1.5%	blunt · at its level
Decision Index 0.4.1: BPoMP (accuracy) submitted	89.1% ✎	▁█▄▄▅█▆▅▄▁▁▁ beats 128/246	easy · avg 83%	▇ 1.3%	sharp · noisy
Decision Index 0.4.1: cfcolor (accuracy) submitted	59.6% ✎	▁········▁▇█ beats 98/246	medium · avg 60%	▇ 0.9%	blunt · at its level
Decision Index 0.4.1: POP909-CL (accuracy) submitted	11.2% ✎	▄▅▅▃▆▅▃▅▂▅▃█ beats 106/237	hard · avg 21%	▇ 0.5%	sharp · noisy
Decision Index 0.4.1: Habermas Machine (accuracy) submitted	47.6% ✎	▁▂▄█▇▅▁▁▄▄▁▁ beats 165/246	medium · avg 47%	▇ 0.3%	blunt · at its level

knowledge & reasoning▁▂▂▂▂▃▃▅▅▆█▂ #141 of 263 · 41% of its score · 14 tests

Beats Winnow-12B (DecisionBench b6ac22b0), Amber Hybrid 4B (decision readout), StartLux-Decision v1.1 2B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: MMLU-Pro (accuracy) submitted	41.8% ✎	▄▄▃▃▂▄▅▇██▅▁ beats 86/246	medium · avg 46%	▇▇▇ 8.1%	sharp · at its level
Decision Index 0.4.1: MMLU (accuracy) submitted	77.8% ✎	▁▄▄▃▃▂▅▇▅█▇▁ beats 137/246	easy · avg 70%	▇▇▇ 6.7%	sharp · at its level
Decision Index 0.4.1: BBH (accuracy) submitted	66.2% ✎	▄▆▄▄█▇▇▄▃▁▁▁ beats 108/246	easy · avg 66%	▇▇ 5.7%	at its level
Decision Index 0.4.1: ARC-Challenge (accuracy) submitted	93.6% ✎	▂▂▂▂▂▂▁▃▄▆█▆ beats 118/246	very easy · avg 90%	▇▇ 5.4%	sharp
Decision Index 0.4.1: ARC-Easy (accuracy) submitted	98.1% ✎	▁▁▂▁▂▂▂▂▃▆█▅ beats 131/246	very easy · avg 96%	▇ 3.3%	sharp · maxed out · noisy
Decision Index 0.4.1: WinoGrande (accuracy) submitted	69.9% ✎	█▅█▅▄█▆▄▃▃▁▁ beats 82/246	easy · avg 78%	▇ 3.0%	typical
Decision Index 0.4.1: GSM8K (accuracy (4- and 10-choice)) submitted	20.4% ✎	▁··▄█▆▆▆▅▅▃▁ beats 39/246	medium · avg 45%	▇ 2.7%	sharp · at its level
Decision Index 0.4.1: GPQA Diamond (accuracy) submitted	48.0% ✎	▂▆▅▅▆█▆▃▁··▁ beats 179/246	medium · avg 39%	▇ 2.4%	blunt · at its level

6 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: MuSR (accuracy) submitted	63.0% ✎	▁▁▂█▅▄▁▁▁▁▁▁ beats 160/246	medium · avg 60%	▇ 1.0%	blunt · at its level
Decision Index 0.4.1: SATA-Bench (case exact accuracy) submitted	29.9% ✎	▁▁▁▁▁▁▁▂▅█▂▁ beats 144/246	hard · avg 22%	▇ 0.9%	typical
Decision Index 0.4.1: CRUXEval (accuracy) submitted	47.2% ✎	██▄▃▃▄▁▁···▁ beats 97/246	medium · avg 58%	▇ 0.9%	at its level · noisy
Decision Index 0.4.1: CLadder (accuracy) submitted	62.3% ✎	▁███▆▅▃▃▁▁▁▁ beats 75/246	easy · avg 71%	▇ 0.8%	at its level
Decision Index 0.4.1: ChessBench (accuracy) submitted	14.0% ✎	▄▇▄▄▄▆▄█▄▁▁▁ beats 104/246	hard · avg 16%	▇ 0.5%	blunt
Decision Index 0.4.1: HLE (accuracy) submitted	10.2% ✎	▁▁▂▃▆█▇▅▅▃▁▁ beats 33/246	very hard · avg 13%	▇ 0.0%	blunt

vision▁·▃▂▃▃▇▄▄█▄▃ #67 of 79 · 10% of its score · 12 tests

Beats StartLux-Decision v1.1 0.8B, mmastrac diffgemma (decision readout) here, though they rank higher overall. Trails LiquidAI d1-omni-600M, Valen-2B, TokenRhythm NeoHorse-Jev-4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index Vision 0.4 · Mind2Web (chance-corrected points) submitted	66.0200 points ✎	▁·▁▁▃▁▄▄▃▅█▂ beats 22/61	avg 63.423 points	▇ 1.2%	blunt
Decision Index Vision 0.4 · MMMU-Pro vision (chance-corrected points) submitted	28.7300 points ✎	▁▁▅▁▄▁▄█▄▃▇▇ beats 23/61	avg 22.797 points	▇ 1.1%	blunt
Decision Index Vision 0.4 · Winoground (chance-corrected points) submitted	50.7500 points ✎	▁·▁▁▃▁·▂▁▃▅█ beats 6/61	avg 66.987 points	▇ 1.0%	blunt
Decision Index Vision 0.4: private same skills (points)	41.0200 points ✎	▁▁▃▃▁▂▄▆▃▃▅█ beats 11/61	avg 48.067 points	▇ 1.0%	blunt
Decision Index Vision 0.4 · Moderation (Hateful Memes) (chance-corrected points) submitted	33.3900 points ✎	▆▃▃▆▄▄▇██▁▆▆ beats 25/61	avg 28.443 points	▇ 0.9%	blunt
Decision Index Vision 0.4 · RealWorldQA (chance-corrected points) submitted	50.6100 points ✎	▁·▂▂▁▃▃▅▅▃▂█ beats 13/61	avg 53.651 points	▇ 0.8%	blunt
Decision Index Vision 0.4 · CharXiv (chance-corrected points) submitted	33.3300 points ✎	▃▁▁▁▁▃▁▅▅▆██ beats 1/61	avg 62.137 points	▇ 0.8%	blunt
Decision Index Vision 0.4 · InfographicVQA (chance-corrected points) submitted	72.6800 points ✎	▁·▁▃▁▁▃▁▂▂▅█ beats 4/61	avg 83.859 points	▇ 0.6%	blunt

4 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index Vision 0.4 · R-Bench-M (chance-corrected points) submitted	23.3000 points ✎	▂▄▇▆█▄▄▇▂··▁ beats 29/61	avg 19.676 points	▇ 0.6%	blunt
Decision Index Vision 0.4 · BLINK (chance-corrected points) submitted	53.7700 points ✎	▁···▁▂▂▂█▄▂▃ beats 33/61	avg 44.994 points	▇ 0.6%	blunt
Decision Index Vision 0.4 · CV-Bench (chance-corrected points) submitted	61.8700 points ✎	▁▁▁▁▃▁▁▁▄█▃▃ beats 6/61	avg 67.662 points	▇ 0.5%	blunt
Decision Index Vision 0.4 · KIE (CORD+FUNSD) (chance-corrected points) submitted	89.6500 points ✎	▁··▁▁▁▁▁▂▄█▁ beats 20/61	avg 85.184 points	▇ 0.3%	blunt

other tests2 not in a task · 6.5% of its score

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: private same skills (equated points)	39.1500 points ✎	▂▂▂▃▃▃▄▅▅▇█▃ beats 95/247	avg 41.206 points	▇▇ 4.2%	blunt
Decision Index 0.4.1: private new domains (equated points)	35.2800 points ✎	▃█▃▃▂▆▃▃▄▄▄▄ beats 104/247	avg 38.861 points	▇ 2.3%	blunt

$ where to run it

Implementation and setup instructions ↗
Published implementation and setup documentation verified. This entry describes a decision readout or evaluated configuration, not necessarily separate trained weights. Follow the referenced version and model settings; hardware compatibility has not been verified. Source · checked 2026-10-31.

Model repository and documentation: github.com/vllm-project ↗

$ how the score was worked out · Decision Index 0.4.1 90%, Decision Index Vision 0.4 10%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
Decision Index 0.4.1 · submitted; independent execution unverified	42	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%	CC BY 4.0
Decision Index Vision 0.4 · submitted; independent execution unverified	12	▇▇▇ 10%	CC BY 4.0
prior · pull toward the middle		▇ 1%
weighted average		= 47.9

not counted (2)

Decision Index 0.4.1	Decision Index 0.4.1: RouterBench (selected quality (quality objective)) [correct]	79.9%	failed run: Publisher excludes this protocol because its prompt reveals GPT-4 as the best route. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json
Decision Index 0.4.1	Decision Index 0.4.1: SGD/SGD-X (macro-F1) [correct]	34.0%	failed run: Publisher reports a builder bug exposing dialogue turns after the classified turn; awaiting corrected rerun. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-11-03 02:37 UTC · sources and licences · json