Winnow-E4B

+ contribute

Winnow-E4B is an open-weights decision model from EldanRing. Its published size is 8B parameters. Licence: apache-2.0.

Catalogue identity notes

Exact Decision Index tested engine winnow-e4b-q8; reported model Winnow-E4B. Type: LoRA; technique: autoregressive. Published checkpoint/base reference: EldanRing/Winnow-E4B. Implementation: https://github.com/EldanRing/winnow-inference. Comparison entry; no current run command, API price or speed inferred from the benchmark. Current repository metadata checked separately; an unpinned alias is not evidence that today’s weights reproduce this historical run.

Model information sources

· Weights licence ↗ · Parameter count ↗

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

46.2±1.1#234 of 355 decision modelsfrom 54 benchmark results across 2 sources

$ how it compares

The decision models ranked either side of Winnow-E4B, out of 357 with a score, then more from EldanRing.

#	model	score	where	RAM
234	Saracura PT-BR 4B Felhen	46±1	model files
235	LatentMesh Blitz 2B v1 LatentMesh Labs	46±1	request access
236	Winnow-E4B EldanRing	46±1	model files
237	Kas 4B kpiya	46±1	model files
238	OpenThai-SystemOne (JevBench mini) iApp Technology	46±2	source code
	more from EldanRing
131	Winnow-12B (DecisionBench b6ac22b0) EldanRing	58±2	local	~16GB
163	Winnow-12B EldanRing	55±1	model files

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

$ what it's good at

tools & automation▁▃▁▁▁▂▃▄▆██▆ #166 of 250 · 7.7% of its score · 6 tests

Beats LiquidAI d1-3B, vLLM-SR Decision 2.0 Nox 4B, Decider 4B here, though they rank higher overall. Trails JevK5 Lite, Gevva E2B, 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: BFCL (case exact accuracy) submitted	91.4% ✎	▁▂▁▁▁▂▂▄▅▆█▁ beats 83/246	very easy · avg 92%	▇ 2.9%	sharp · noisy
Decision Index 0.4.1: API-Bank (accuracy) submitted	73.0% ✎	▁▁▁▁▁▁▁▂▃▃█▂ beats 98/236	easy · avg 70%	▇ 2.3%	sharp · at its level · noisy
Decision Index 0.4.1: When2Call (accuracy) submitted	85.0% ✎	▁▁▁▂▃▅▅▆█▇█▃ beats 172/246	easy · avg 72%	▇ 1.0%	at its level · noisy
Decision Index 0.4.1: ToolRet (nDCG@10) submitted	60.3% ✎	▁▁▁▁▁·▁▁▁▄█▁ beats 66/246	medium · avg 60%	▇ 0.8%	blunt · at its level
Decision Index 0.4.1: Home appliance simulator (case exact accuracy) submitted	10.2% ✎	█▁▁▃▄▄▆▄▁▂▁▃ beats 69/243	medium · avg 37%	▇ 0.7%	sharp · at its level · noisy
Decision Index 0.4.1: RouterBench (selected quality (quality objective)) submitted	79.8% ✎	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▁▁▁·▁▁▁▁▃▃█▅ #166 of 258 · 5.2% of its score · 7 tests

Beats Solar Decide, decisio 0.4.0 (Qwen3.6-35B-A3B) (decision readout), JevK5 9B v0.3.3 here, though they rank higher overall. Trails Bosun v3.1 1.7B LoRA, Dinah-1, DecisionTune 1.0 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	41.5% ✎	▁▁▁▁▁▁▁▂▂▄▇█ beats 111/246	medium · avg 38%	▇ 1.7%	blunt · at its level
Decision Index 0.4.1: Amazon ESCI (macro-F1) submitted	46.6% ✎	▁▁▁▁▁▁▂▂▃▃█▅ beats 90/246	medium · avg 47%	▇ 1.7%	at its level
Decision Index 0.4.1: HoVer (accuracy) submitted	64.8% ✎	▆▆▇▅▄▄▃▇▄█▆▄ beats 80/246	easy · avg 74%	▇ 0.5%	blunt · at its level
Decision Index 0.4.1: PhishNChips phishing decisions (accuracy) submitted	50.9% ✎	█▅▅▄▃▂▁▁▁▁·▁ beats 46/245	easy · avg 70%	▇ 0.5%	at its level · noisy
Decision Index 0.4.1: CLINC150+OOS (macro-F1) submitted	82.6% ✎	▁▁▁▁▁▁▁▂▃▆█▄ beats 74/238	easy · avg 83%	▇ 0.4%	sharp · noisy
Decision Index 0.4.1: BANKING77 (macro-F1) submitted	71.2% ✎	▁·▁▁▁▁▁▁▂█▇█ beats 42/238	easy · avg 79%	▇ 0.3%	noisy
Decision Index 0.4.1: SGD/SGD-X (macro-F1) submitted	37.3% ✎	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▂▂▂▃▂▅▆▃█▆▄▃ #183 of 256 · 22% of its score · 9 tests

Beats Nari Decision 4B v3.2, Solar Decide Flash, Lev1 3.7B here, though they rank higher overall. Trails DecisionTune 1.0, vera-spark, jdi-ptr14-0.8b 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	76.6% ✎	▂▄▂▂▄▇▃▇█▅▃▁ beats 67/246	very easy · avg 85%	▇▇▇ 6.8%	sharp
Decision Index 0.4.1: NLI4CT (macro-F1) submitted	73.9% ✎	▁▁▁▂▂▃▃▄▇▇█▄ beats 92/246	easy · avg 74%	▇▇ 4.9%	at its level
Decision Index 0.4.1: ANLI (macro-F1) submitted	56.2% ✎	▁▂▃▅▅█▂····▁ beats 89/246	medium · avg 60%	▇▇ 4.4%	at its level
Decision Index 0.4.1: iSarcasmEval (Sarcasm F1 · track A, English) submitted	53.8% ✎	▁·▁·▁▁▂▄▅█▇▄ beats 150/246	medium · avg 44%	▇ 1.7%	at its level
Decision Index 0.4.1: ContractNLI (macro-F1) submitted	73.3% ✎	▁▁▁▁▁▁▁▂▂▇▇█ beats 98/246	easy · avg 71%	▇ 1.0%	at its level · noisy
Decision Index 0.4.1: VAST (macro-F1) submitted	63.1% ✎	▁▁▁▂▅▄▄▃▄▆█▄ beats 114/246	medium · avg 62%	▇ 0.8%	at its level
Decision Index 0.4.1: FinEntity (macro-F1) submitted	88.5% ✎	▁···▁▁▁▁▃▆█▃ beats 130/246	easy · avg 84%	▇ 0.7%	noisy
Decision Index 0.4.1: RAGTruth (F1 on hallucinated class) submitted	60.8% ✎	▁▁▁▁▁▁▁▁▁▃▆█ beats 68/246	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: ACOS (per-review F1) submitted	20.6% ✎	▁··▂▃▃▅▇█▇▃▃ beats 111/246	hard · avg 22%	▇ 0.7%	noisy

arts & human taste▁▇▄▃▅▅█▅▄▂▁▁ #155 of 247 · 7.7% of its score · 6 tests

Beats Jegue 0.1 (9B), Standard One 8B SH (v4), Sieve 4B here, though they rank higher overall. Trails Crivo-32M v2, Crivo-68M, Crivo-150M v2 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	55.7% ✎	▁▁▂▃▃▃▇▅█▅▃▁ beats 76/246	medium · avg 61%	▇ 3.4%	at its level
Decision Index 0.4.1: Humicroedit (accuracy) submitted	60.3% ✎	▁·▁▃▂▂▅█▄▂▁▁ beats 100/246	medium · avg 60%	▇ 1.5%	blunt · at its level
Decision Index 0.4.1: BPoMP (accuracy) submitted	81.5% ✎	▁█▄▄▅█▆▅▄▁▁▁ beats 93/246	easy · avg 83%	▇ 1.3%	sharp · noisy
Decision Index 0.4.1: cfcolor (accuracy) submitted	56.6% ✎	▁········▁▇█ beats 63/246	medium · avg 60%	▇ 0.8%	blunt · at its level
Decision Index 0.4.1: POP909-CL (accuracy) submitted	3.6% ✎	▄▅▅▃▆▅▃▅▂▅▃█ beats 66/237	hard · avg 21%	▇ 0.4%	sharp · noisy
Decision Index 0.4.1: Habermas Machine (accuracy) submitted	44.2% ✎	▁▂▄█▇▅▁▁▄▄▁▁ beats 122/246	medium · avg 47%	▇ 0.2%	blunt · at its level

knowledge & reasoning▁▂▂▂▂▃▃▅▅▆█▂ #175 of 263 · 42% of its score · 14 tests

Beats baselabs joey-1 (2B), LiquidAI d1-3B, Aurelis-2B here, though they rank higher overall. Trails openvons Qwen3-4B (decision readout), Jevstral 8B, Jobe Qwen3.5-4B (decision readout) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: MMLU-Pro (accuracy) submitted	43.8% ✎	▄▄▃▃▂▄▅▇██▅▁ beats 94/246	medium · avg 46%	▇▇▇ 7.7%	sharp · at its level
Decision Index 0.4.1: MMLU (accuracy) submitted	67.0% ✎	▁▄▄▃▃▂▅▇▅█▇▁ beats 84/246	easy · avg 70%	▇▇▇ 6.7%	sharp · at its level
Decision Index 0.4.1: ARC-Challenge (accuracy) submitted	88.6% ✎	▂▂▂▂▂▂▁▃▄▆█▆ beats 86/246	very easy · avg 90%	▇▇ 6.0%	sharp
Decision Index 0.4.1: BBH (accuracy) submitted	62.9% ✎	▄▆▄▄█▇▇▄▃▁▁▁ beats 95/246	easy · avg 66%	▇▇ 5.7%	at its level
Decision Index 0.4.1: ARC-Easy (accuracy) submitted	96.0% ✎	▁▁▂▁▂▂▂▂▃▆█▅ beats 90/246	very easy · avg 96%	▇▇ 3.8%	sharp · maxed out · noisy
Decision Index 0.4.1: WinoGrande (accuracy) submitted	66.1% ✎	█▅█▅▄█▆▄▃▃▁▁ beats 62/246	easy · avg 78%	▇ 3.1%	at its level
Decision Index 0.4.1: GSM8K (accuracy (4- and 10-choice)) submitted	26.4% ✎	▁··▄█▆▆▆▅▅▃▁ beats 63/246	medium · avg 45%	▇ 2.5%	sharp · at its level
Decision Index 0.4.1: GPQA Diamond (accuracy) submitted	36.2% ✎	▂▆▅▅▆█▆▃▁··▁ beats 90/246	medium · avg 39%	▇ 2.3%	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	55.2% ✎	▁▁▂█▅▄▁▁▁▁▁▁ beats 72/246	medium · avg 60%	▇ 1.0%	blunt · at its level
Decision Index 0.4.1: CRUXEval (accuracy) submitted	50.2% ✎	██▄▃▃▄▁▁···▁ beats 112/246	medium · avg 58%	▇ 0.9%	at its level · noisy
Decision Index 0.4.1: SATA-Bench (case exact accuracy) submitted	29.2% ✎	▁▁▁▁▁▁▁▂▅█▂▁ beats 133/246	hard · avg 22%	▇ 0.8%	typical
Decision Index 0.4.1: CLadder (accuracy) submitted	60.3% ✎	▁███▆▅▃▃▁▁▁▁ beats 65/246	easy · avg 71%	▇ 0.8%	at its level
Decision Index 0.4.1: ChessBench (accuracy) submitted	15.0% ✎	▄▇▄▄▄▆▄█▄▁▁▁ beats 114/246	hard · avg 16%	▇ 0.4%	blunt
Decision Index 0.4.1: HLE (accuracy) submitted	12.6% ✎	▁▁▂▃▆█▇▅▅▃▁▁ beats 133/246	very hard · avg 13%	▇ 0.0%	blunt

vision▁·▃▂▃▃▇▄▄█▄▃ #73 of 79 · 9.3% of its score · 12 tests

Trails LiquidAI d1-omni-600M, InternLM Intern-Decision Qwen3.5-2B, Valen-2B 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	53.2700 points ✎	▁·▁▁▃▃▄▄▃▅█▂ beats 9/61	avg 63.423 points	▇ 1.2%	blunt
Decision Index Vision 0.4 · MMMU-Pro vision (chance-corrected points) submitted	11.1600 points ✎	▁▁▅▃▄▁▄█▄▃▇▇ beats 9/61	avg 22.797 points	▇ 1.1%	blunt
Decision Index Vision 0.4 · Winoground (chance-corrected points) submitted	63.2500 points ✎	▁·▁▁▁▁·▃▁▃▅█ beats 12/61	avg 66.987 points	▇ 1.0%	blunt
Decision Index Vision 0.4: private same skills (points)	36.0600 points ✎	▁▁▃▃▁▂▄▆▃▃▅█ beats 3/61	avg 48.067 points	▇ 1.0%	blunt
Decision Index Vision 0.4 · Moderation (Hateful Memes) (chance-corrected points) submitted	15.9100 points ✎	▆▃▃▆▄▄▇██▁▆▆ beats 8/61	avg 28.443 points	▇ 0.9%	blunt
Decision Index Vision 0.4 · RealWorldQA (chance-corrected points) submitted	35.0800 points ✎	▁·▃▂▁▃▃▅▅▃▂█ beats 1/61	avg 53.651 points	▇ 0.8%	blunt
Decision Index Vision 0.4 · CharXiv (chance-corrected points) submitted	43.3600 points ✎	▂▁▃▁▁▃▁▅▅▆██ beats 4/61	avg 62.137 points	▇ 0.8%	blunt
Decision Index Vision 0.4 · InfographicVQA (chance-corrected points) submitted	83.0600 points ✎	▁·▁▂▁▁▃▁▂▂▅█ beats 16/61	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	15.1800 points ✎	▂▄▇▆█▄▄▇▂··▁ beats 9/61	avg 19.676 points	▇ 0.6%	blunt
Decision Index Vision 0.4 · BLINK (chance-corrected points) submitted	34.7600 points ✎	▁···▁▃▂▂█▄▂▃ beats 8/61	avg 44.994 points	▇ 0.6%	blunt
Decision Index Vision 0.4 · CV-Bench (chance-corrected points) submitted	49.4900 points ✎	▃▁▁▁▂▁▁▁▄█▃▃ beats 0/61	avg 67.662 points	▇ 0.5%	blunt
Decision Index Vision 0.4 · KIE (CORD+FUNSD) (chance-corrected points) submitted	80.0300 points ✎	▁··▁▁▁▃▁▂▄█▁ beats 6/61	avg 85.184 points	▇ 0.3%	blunt

other tests2 not in a task · 6.3% of its score

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: private same skills (equated points)	36.1700 points ✎	▂▂▂▃▃▃▄▅▅▇█▃ beats 82/247	avg 41.206 points	▇▇ 4.1%	blunt
Decision Index 0.4.1: private new domains (equated points)	27.5400 points ✎	▃█▃▃▂▆▃▃▄▄▄▄ beats 93/247	avg 38.861 points	▇ 2.2%	blunt

$ where to run it

Download model files ↗
Published model files verified. This is a verified repository snapshot; the benchmark does not establish an identical weight revision. A local run command and hardware requirements have not been verified for this entry. Source · checked 2026-10-31.

Model repository and documentation: huggingface.co/EldanRing ↗

$ how the score was worked out · Decision Index 0.4.1 90%, Decision Index Vision 0.4 9%

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	▇▇▇ 9%	CC BY 4.0
prior · pull toward the middle		▇ 1%
weighted average		= 46.2

not counted (2)

Decision Index 0.4.1	Decision Index 0.4.1: RouterBench (selected quality (quality objective)) [correct]	79.8%	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]	37.3%	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