whatmodel.sh / decision / Nace.AI Drex v2.0 (35B-A3B)

Nace.AI Drex v2.0 (35B-A3B)

+ contribute

Nace.AI Drex v2.0 (35B-A3B) is an open-weights decision model from Nace.AI. Its published size is 35.11B parameters. Licence: other.

Catalogue identity notes

Exact Decision Index tested engine drex-v2.0; reported model Nace.AI Drex v2.0 (35B-A3B). Type: full fine-tune; technique: autoregressive. 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.

62.9±1.1#78 of 355 decision modelsfrom 67 benchmark results across 4 sources

$ how it compares

The decision models ranked either side of Nace.AI Drex v2.0 (35B-A3B), out of 357 with a score, then more from Nace.AI.

#	model	score	where	released
76	Microsoft-Decision-1 Microsoft	63±1	api price unknown	2026-10-09
77	Mercury Decide (free endpoint) Inception	63±2	api price unknown
78	Nace.AI Drex v2.0 (35B-A3B) Nace.AI	63±1	model files
79	Sieve-Vision-27B Sthānika AI	63±4	model files
80	reflex-27b (DecisionBench 1d4bf0f2) kshetrajna12	63±2	model files
	more from Nace.AI
133	Nace.AI Drex v1.5 Nace.AI	58±1	model files

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

$ what it's good at

tools & automation▁▃▁▁▁▂▃▄▆██▆ #75 of 250 · 6.1% of its score · 8 tests

Beats Sieve 27B v2, allec-oavg (Gemma 4 31B), Jev 1.13 here, though they rank higher overall. Trails baselabs joey-1 (2B), Hopper 12B (trained), Solomon v1.1 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: API-Bank (accuracy) submitted	83.3% ✎	▁▁▁▁▁▁▁▂▃▂█▂ beats 135/236	easy · avg 70%	▇ 1.2%	sharp · noisy
DecideBench v1: agent routing	98.0% ✎	▄·····▂▄·█·▄ beats 8/10	very easy · avg 93%	▇ 1.2%	sharp · maxed out · noisy
DecideBench v1: action review	98.0% ✎	▃▃·▃·█···█·▅ beats 7/10	easy · avg 87%	▇ 1.0%	sharp · maxed out · noisy
Decision Index 0.4.1: BFCL (case exact accuracy) submitted	96.6% ✎	▁▂▁▁▁▂▂▄▅▆█▁ beats 170/246	very easy · avg 92%	▇ 0.7%	sharp · maxed out · noisy
Decision Index 0.4.1: Home appliance simulator (case exact accuracy) submitted	54.5% ✎	█▁▁▃▄▄▆▄▁▂▁▃ beats 139/243	medium · avg 37%	▇ 0.7%	sharp · at its level · noisy
Decision Index 0.4.1: ToolRet (nDCG@10) submitted	67.3% ✎	▁▁▁▁▁·▁▁▁▄█▁ beats 186/246	medium · avg 60%	▇ 0.7%	blunt · at its level
Decision Index 0.4.1: When2Call (accuracy) submitted	86.6% ✎	▁▁▁▂▃▅▅▆█▇█▃ beats 184/246	easy · avg 72%	▇ 0.6%	noisy
Decision Index 0.4.1: RouterBench (selected quality (quality objective)) submitted	79.2% ✎	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▁▁▁·▁▁▁▁▃▃█▅ #110 of 258 · 12% of its score · 14 tests

Beats AutoTrust-JEV-27B v1.1, Quyet-1.0-Large, OneJev1.6-9B here, though they rank higher overall. Trails vLLM-SR vela3-fast (Vela 3.0, 0.4B), crystal-0.8b, Jev-Style 2B v4 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
System One Bench: MIRACL pairwise (Arabic)	89.0% ✎	▂···▁··▃▄▂█▇ beats 14/29	easy · avg 84%	▇ 2.5%	typical
System One Bench: MASSIVE intent (Arabic)	82.5% ✎	▄·▁·▁▃▃▃█▄▇▇ beats 10/29	easy · avg 76%	▇ 2.2%	typical
Decision Index 0.4.1: BRIGHT (nDCG@10) submitted	47.2% ✎	▁▁▁▁▁▁▁▂▂▄▇█ beats 188/246	medium · avg 38%	▇ 1.8%	blunt · at its level
Decision Index 0.4.1: Amazon ESCI (macro-F1) submitted	54.6% ✎	▁▁▁▁▁▁▂▂▃▃█▅ beats 147/246	medium · avg 47%	▇ 1.6%	at its level
System One Bench: MIRACL rerank (Arabic)	87.0% ✎	▁▂·▁▁·▂·▄▄█▇ beats 15/29	easy · avg 79%	▇ 1.4%	typical
System One Bench: MASSIVE scenario (Arabic)	63.5% ✎	▃▁·▄▃▂█▂▁▁·▃ beats 11/29	medium · avg 60%	▇ 1.0%	at its level
DecideBench v1: support intent	92.0% ✎	▃····▃▂·▃··█ beats 2/10	very easy · avg 95%	▇ 0.4%	sharp · maxed out · noisy
Decision Index 0.4.1: HoVer (accuracy) submitted	85.8% ✎	▆▆▇▅▄▄▃▇▄█▆▄ beats 189/246	easy · avg 74%	▇ 0.4%	blunt

6 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: PhishNChips phishing decisions (accuracy) submitted	75.0% ✎	█▅▅▄▃▂▁▁▁▁·▁ beats 151/245	easy · avg 70%	▇ 0.3%	noisy
System One Bench: AG News (English)	88.0% ✎	▁···▅▁▅▂▆█▂▅ beats 9/29	easy · avg 88%	▇ 0.3%	blunt
Decision Index 0.4.1: CLINC150+OOS (macro-F1) submitted	89.1% ✎	▁▁▁▁▁▁▁▂▃▆█▄ beats 111/238	easy · avg 83%	▇ 0.2%	sharp · noisy
Decision Index 0.4.1: BANKING77 (macro-F1) submitted	80.5% ✎	▁·▁▁▁▁▁▁▂█▇█ beats 103/238	easy · avg 79%	▇ 0.2%	noisy
System One Bench: Banking77 (English)	80.5% ✎	█▅▆█▅▃▃▃···▂ beats 4/26	very easy · avg 87%	▇ 0.0%	blunt · noisy
Decision Index 0.4.1: SGD/SGD-X (macro-F1) submitted	47.4% ✎	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▂▂▂▃▂▅▆▃█▆▄▃ #45 of 256 · 25% of its score · 21 tests

Beats Ainvest-Decision-27B v1.0, Turion Jev 31B, MISA AI 2.0 Decision Plus · oak 1.1 (31B) here, though they rank higher overall. Trails Decider 35B-A3B, Sieve 9B v2, Clef-flash here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: ANLI (macro-F1) submitted	70.5% ✎	▁▂▃▅▅█▂····▁ beats 164/246	medium · avg 60%	▇ 3.4%	at its level
Decision Index 0.4.1: NLI4CT (macro-F1) submitted	81.9% ✎	▁▁▁▂▂▃▃▄▇▇█▄ beats 168/246	easy · avg 74%	▇ 3.3%	typical
DecideBench v1: review sentiment	94.0% ✎	▂··▂···▂▂█▄▂ beats 7/10	easy · avg 85%	▇ 2.2%	sharp · maxed out
Decision Index 0.4.1: HellaSwag (accuracy) submitted	95.9% ✎	▂▄▂▂▄▇▃▇█▅▃▁ beats 206/246	very easy · avg 85%	▇ 2.2%	sharp · maxed out
System One Bench: BoolQ (English)	92.5% ✎	▁▁▁▃·▁▃▃▇▇█▅ beats 28/29	easy · avg 86%	▇ 1.7%	typical
Decision Index 0.4.1: iSarcasmEval (Sarcasm F1 · track A, English) submitted	63.7% ✎	▁·▁·▁▁▂▄▅█▇▄ beats 203/246	medium · avg 44%	▇ 1.7%	at its level
System One Bench: OSACT4 offensive (Arabic)	94.5% ✎	▁······▁▃█▇▄ beats 26/29	easy · avg 84%	▇ 1.6%	sharp
DecideBench v1: content moderation	98.0% ✎	▂··▂···▄▄▂·█ beats 10/10	easy · avg 83%	▇ 1.5%	sharp · maxed out · noisy

13 more tests

test	result	vs the others in its run	difficulty	counts	why
System One Bench: SST-2 (English)	94.0% ✎	▁·▃▁▂▂▃▅█·▂▁ beats 11/29	easy · avg 92%	▇ 1.0%	typical
Decision Index 0.4.1: ACOS (per-review F1) submitted	23.9% ✎	▁··▂▃▃▅▇█▇▃▃ beats 133/246	hard · avg 22%	▇ 1.0%	at its level · noisy
System One Bench: SST-5 (English)	56.0% ✎	▃·▁▁▃·▇▄▄▄█▅ beats 16/29	medium · avg 51%	▇ 1.0%	blunt · at its level
DecideBench v1: claim support	96.0% ✎	▄··▄·▄·▄▄███ beats 8/10	easy · avg 82%	▇ 0.7%	sharp · noisy
Decision Index 0.4.1: VAST (macro-F1) submitted	78.4% ✎	▁▁▁▂▅▄▄▂▄▆█▄ beats 202/246	medium · avg 62%	▇ 0.7%	at its level
Decision Index 0.4.1: ContractNLI (macro-F1) submitted	83.4% ✎	▁▁▁▁▁▁▁▂▂▇▇█ beats 200/246	easy · avg 71%	▇ 0.6%	noisy
System One Bench: XNLI Arabic (Arabic)	71.3% ✎	▄▁▃▁▄▇█▅▃▁·▃ beats 10/29	medium · avg 68%	▇ 0.5%	at its level
Decision Index 0.4.1: RAGTruth (F1 on hallucinated class) submitted	82.5% ✎	▁▁▁▁▁▁▁▁▁▃▆█ beats 206/246	easy · avg 66%	▇ 0.5%	noisy
System One Bench: XNLI French (French)	77.3% ✎	▂▅·▁▁▅▅█▃▁·▃ beats 10/29	medium · avg 76%	▇ 0.4%	blunt · at its level
System One Bench: Tweet sentiment (Arabic)	69.5% ✎	▁······▁▁▄█▄ beats 18/29	medium · avg 62%	▇ 0.4%	blunt · at its level
Decision Index 0.4.1: FinEntity (macro-F1) submitted	90.8% ✎	▁···▁▁▁▁▃▆█▃ beats 162/246	easy · avg 84%	▇ 0.3%	noisy
System One Bench: XNLI English (English)	87.3% ✎	▃·▁·▃▇█▄█▃▃▁ beats 15/29	easy · avg 85%	▇ 0.1%	blunt · noisy
System One Bench: DAIR Emotion (English)	55.5% ✎	██▃·····▁··▁ beats 6/29	medium · avg 59%	▇ 0.0%	blunt · at its level

arts & human taste▁▇▄▃▅▅█▅▄▂▁▁ #118 of 247 · 6.2% of its score · 6 tests

Beats Bobcat 1.3 (31B) 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	77.6% ✎	▁▁▂▃▃▃▇▅█▅▃▁ beats 206/246	medium · avg 61%	▇ 2.5%	at its level
Decision Index 0.4.1: Humicroedit (accuracy) submitted	63.5% ✎	▁·▁▃▂▂▅█▄▂▁▁ beats 183/246	medium · avg 60%	▇ 1.4%	blunt · at its level
Decision Index 0.4.1: POP909-CL (accuracy) submitted	79.8% ✎	▄▅▅▃▆▅▃▅▂▅▃█ beats 218/237	hard · avg 21%	▇ 0.8%	sharp · at its level · noisy
Decision Index 0.4.1: cfcolor (accuracy) submitted	59.2% ✎	▁········▁▇█ beats 95/246	medium · avg 60%	▇ 0.8%	blunt · at its level
Decision Index 0.4.1: BPoMP (accuracy) submitted	89.0% ✎	▁█▄▄▅█▆▅▄▁▁▁ beats 127/246	easy · avg 83%	▇ 0.6%	sharp · noisy
Decision Index 0.4.1: Habermas Machine (accuracy) submitted	62.4% ✎	▁▂▄█▇▅▁▁▄▄▁▁ beats 224/246	medium · avg 47%	▇ 0.2%	blunt · at its level

knowledge & reasoning▁▂▂▂▂▃▃▅▅▆█▂ #63 of 263 · 39% of its score · 15 tests

Beats StartLux-Decision v1.1 9B, Sieve 27B v2, Standard One 35B SH (v1) here, though they rank higher overall. Trails Solar Decide, Clef-flash, Clef-omni here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Etsabary · 1342 text decisions	68.2% ✎	▁··▂·▃▅▁▃▂▂█ beats 6/24	medium · avg 47%	▇▇▇▇▇ 12%	sharp · at its level
Decision Index 0.4.1: MMLU-Pro (accuracy) submitted	72.8% ✎	▄▄▃▃▂▄▅▇██▅▁ beats 224/246	medium · avg 46%	▇▇▇ 7.0%	sharp · at its level
Decision Index 0.4.1: MMLU (accuracy) submitted	85.3% ✎	▁▄▄▃▃▂▅▇▅█▇▁ beats 193/246	easy · avg 70%	▇▇ 3.9%	sharp
Decision Index 0.4.1: BBH (accuracy) submitted	78.3% ✎	▄▆▄▄█▇▇▄▃▁▁▁ beats 180/246	easy · avg 66%	▇▇ 3.8%	typical
Decision Index 0.4.1: GPQA Diamond (accuracy) submitted	51.5% ✎	▂▆▅▅▆█▆▃▁··▁ beats 208/246	medium · avg 39%	▇ 2.4%	blunt · at its level
Decision Index 0.4.1: GSM8K (accuracy (4- and 10-choice)) submitted	66.8% ✎	▁··▄█▆▆▆▅▅▃▁ beats 179/246	medium · avg 45%	▇ 2.3%	sharp · at its level
Decision Index 0.4.1: WinoGrande (accuracy) submitted	86.1% ✎	█▅█▅▄█▆▄▃▃▁▁ beats 170/246	easy · avg 78%	▇ 1.8%	typical
Decision Index 0.4.1: ARC-Challenge (accuracy) submitted	96.8% ✎	▂▂▂▂▂▂▁▃▄▆█▆ beats 184/246	very easy · avg 90%	▇ 1.4%	sharp · maxed out

7 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: SATA-Bench (case exact accuracy) submitted	24.1% ✎	▁▁▁▁▁▁▁▂▅█▂▁ beats 101/246	hard · avg 22%	▇ 1.2%	at its level
Decision Index 0.4.1: MuSR (accuracy) submitted	59.6% ✎	▁▁▂█▅▄▁▁▁▁▁▁ beats 128/246	medium · avg 60%	▇ 0.8%	blunt · at its level
Decision Index 0.4.1: CRUXEval (accuracy) submitted	84.6% ✎	██▄▃▃▄▁▁···▁ beats 211/246	medium · avg 58%	▇ 0.7%	at its level · noisy
Decision Index 0.4.1: ARC-Easy (accuracy) submitted	98.6% ✎	▁▁▂▁▂▂▂▂▃▆█▅ beats 159/246	very easy · avg 96%	▇ 0.6%	sharp · maxed out · noisy
Decision Index 0.4.1: CLadder (accuracy) submitted	84.5% ✎	▁███▆▅▃▃▁▁▁▁ beats 202/246	easy · avg 71%	▇ 0.6%	typical
Decision Index 0.4.1: ChessBench (accuracy) submitted	23.4% ✎	▄▇▄▄▄▆▄█▄▁▁▁ beats 208/246	hard · avg 16%	▇ 0.5%	blunt
Decision Index 0.4.1: HLE (accuracy) submitted	12.8% ✎	▁▁▂▃▆█▇▅▅▃▁▁ beats 141/246	very hard · avg 13%	▇ 0.0%	blunt

workflows▁▁▁▂▁▂▂▆█▄▁▁ #16 of 114 · 5.3% of its score · 3 tests

Beats Nace.AI Drex Max v1 (27B), d1-ultra, baselabs elephant-1 (9B) here, though they rank higher overall. Trails GPT-6 Luna Decisions, reflex-27b (DecisionBench 1d4bf0f2) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
System One Bench: Typed decisions (English)	68.7% ✎	▁▁·▁▁▁▅▄▇██▁ beats 12/29	medium · avg 63%	▇ 2.3%	blunt · at its level
DecideBench v1: ticket triage	100.0% ✎	▄·····▂█·▆·▃ beats 10/10	very easy · avg 90%	▇ 1.9%	sharp · maxed out
DecideBench v1: returns policy	94.0% ✎	▃··▃··▃·▅▅█▃ beats 9/10	easy · avg 81%	▇ 1.1%	sharp · noisy

other tests2 not in a task · 6.1% of its score

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: private same skills (equated points)	53.8700 points ✎	▂▂▂▃▃▃▄▅▅▇█▃ beats 164/247	avg 41.206 points	▇▇ 3.9%	blunt
Decision Index 0.4.1: private new domains (equated points)	51.0800 points ✎	▃█▃▃▂▆▃▃▄▄▄▄ beats 160/247	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/nace-ai ↗

$ how the score was worked out · Decision Index 0.4.1 62%, System One Bench (Ali Alhashi… 16%, Etsabary · decision benchmarks 12% and 1 more

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	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 62%	CC BY 4.0
System One Bench (Ali Alhashimi)	16	▇▇▇▇▇ 16%	CC-BY-4.0
Etsabary · decision benchmarks	1	▇▇▇ 12%	MIT
DecideBench (Cho Yin Yong)	8	▇▇▇ 10%	CC BY 4.0
prior · pull toward the middle		▇ 1%
weighted average		= 62.9

not counted (2)

Decision Index 0.4.1	Decision Index 0.4.1: RouterBench (selected quality (quality objective)) [correct]	79.2%	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]	47.4%	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