Decider 4B

+ contribute

Decider 4B is an open-weights decision model from Mapika. Its published size is 4.66B parameters. Licence: apache-2.0.

Catalogue identity notes

Exact Decision Index tested engine decider-4b; reported model Decider 4B. Type: full fine-tune; technique: autoregressive. Implementation: https://github.com/Mapika/decider. No 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. Reviewed Ollaya local route added 2026-10-31: decider:4b, manifest b4e1276b55eda06fad7af9536a996fc132a1229bb866755cd6f07778dbe98963, matching the independent accuracy-run manifest. Native family implementation; published runtime precision label F32. Memory is measured peak GPU increase (11279 MiB, rounded upward to 11.1 GiB) on RTX 5090 during the short five-question triage workload, not download size or a maximum-context guarantee. Runtime installation is required; no fresh execution performed here. Historical benchmark adapter and input protocol still vary by result.

Model information sources

· Weights licence ↗ · Parameter count ↗

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

50.6±1.1#199 of 355 decision modelsfrom 57 benchmark results across 3 sources

$ how it compares

The decision models ranked either side of Decider 4B, out of 357 with a score, then more from Mapika.

#	model	score	where	RAM	released
199	djev-spark (DecisionBench 1444f3e9) mmastrac	51±2	source code
200	vLLM-SR Decision 2.0 Nox 4B vLLM Semantic Router	51±1	local	~12GB
201	Decider 4B Mapika	51±1	local	~11GB
202	Together Tev1 Qwen3.5-4B Together	51±1	model files
203	OneJev-4B (JevBench mini) OmniJev	51±2	source code
	more from Mapika
43	Decider 31B (NVFP4) Mapika	66±1	model files		2026-10
85	Decider chat · Gemma-4-31B (decision readout) Mapika	62±1	source code
110	decider-35b-a3b (DecisionBench d9783c58) Mapika	60±2	model files		2026-09-20

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

$ what it's good at

tools & automation▁▃▁▁▁▂▃▄▆██▆ #183 of 250 · 6.6% of its score · 6 tests

Trails JevK5 Lite, Mola-68M, Mola-150M 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	88.0% ✎	▁▂▁▁▁▂▃▄▅▆█▁ beats 66/246	very easy · avg 92%	▇ 2.1%	sharp · maxed out · noisy
Decision Index 0.4.1: API-Bank (accuracy) submitted	60.4% ✎	▁▁▁▁▁▁▁▂▃▂█▂ beats 71/236	easy · avg 70%	▇ 2.0%	sharp · at its level · noisy
Decision Index 0.4.1: When2Call (accuracy) submitted	75.9% ✎	▁▁▁▂▃▅▅▆█▇█▃ beats 117/246	easy · avg 72%	▇ 0.9%	at its level · noisy
Decision Index 0.4.1: Home appliance simulator (case exact accuracy) submitted	17.1% ✎	█▁▁▃▄▄▆▄▁▂▁▃ beats 82/243	medium · avg 37%	▇ 0.8%	sharp · at its level · noisy
Decision Index 0.4.1: ToolRet (nDCG@10) submitted	65.0% ✎	▁▁▁▁▁·▁▁▁▄█▁ beats 132/246	medium · avg 60%	▇ 0.8%	blunt · at its level
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▁▁▁·▁▁▁▁▃▃█▅ #183 of 258 · 5.7% of its score · 8 tests

Beats JevK5 9B v0.3.3 here, though they rank higher overall. Trails Mola-68M, Bosun v3.1 1.7B LoRA, Dinah-1 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	45.5% ✎	▁▁▁▁▁▁▁▂▂▄▇█ beats 161/246	medium · avg 38%	▇ 1.8%	blunt · at its level
Decision Index 0.4.1: Amazon ESCI (macro-F1) submitted	42.4% ✎	▁▁▁▁▁▁▂▂▃▃█▅ beats 71/246	medium · avg 47%	▇ 1.7%	at its level
GEO Citation · listwise Spearman (October 2026)	0.4803 spearman-rho ✎	▃···▃······█ beats 0/4	avg 0.256 spearman-rho	▇ 0.7%	blunt
Decision Index 0.4.1: HoVer (accuracy) submitted	56.3% ✎	▆▆▇▅▄▄▃▇▄█▆▄ beats 28/246	easy · avg 74%	▇ 0.5%	blunt · at its level
Decision Index 0.4.1: PhishNChips phishing decisions (accuracy) submitted	57.8% ✎	█▅▅▄▃▂▁▁▁▁·▁ beats 79/245	easy · avg 70%	▇ 0.4%	at its level · noisy
Decision Index 0.4.1: CLINC150+OOS (macro-F1) submitted	86.6% ✎	▁▁▁▁▁▁▁▂▃▆█▄ beats 87/238	easy · avg 83%	▇ 0.4%	sharp · noisy
Decision Index 0.4.1: BANKING77 (macro-F1) submitted	74.7% ✎	▁·▁▁▁▁▁▁▂█▇█ beats 59/238	easy · avg 79%	▇ 0.3%	noisy
Decision Index 0.4.1: SGD/SGD-X (macro-F1) submitted	20.0% ✎	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▂▂▂▃▂▅▆▃█▆▄▃ #87 of 256 · 19% of its score · 9 tests

Beats Turion Jev 31B, AJev lm6q (Gemma 4 31B), oenogarum (Gemma 4 31B) here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: HellaSwag (accuracy) submitted	94.0% ✎	▂▄▂▂▄▇▃▇█▅▃▁ beats 174/246	very easy · avg 85%	▇▇ 5.3%	sharp
Decision Index 0.4.1: NLI4CT (macro-F1) submitted	73.7% ✎	▁▁▁▂▂▃▃▄▇▇█▄ beats 90/246	easy · avg 74%	▇▇ 4.5%	at its level
Decision Index 0.4.1: ANLI (macro-F1) submitted	56.4% ✎	▁▂▃▅▅█▂····▁ beats 92/246	medium · avg 60%	▇▇ 4.2%	at its level
Decision Index 0.4.1: iSarcasmEval (Sarcasm F1 · track A, English) submitted	42.1% ✎	▁·▁·▁▁▂▄▅█▇▄ beats 98/246	medium · avg 44%	▇ 1.7%	at its level
Decision Index 0.4.1: ContractNLI (macro-F1) submitted	72.3% ✎	▁▁▁▁▁▁▁▂▂▇▇█ beats 88/246	easy · avg 71%	▇ 0.9%	at its level · noisy
Decision Index 0.4.1: VAST (macro-F1) submitted	53.5% ✎	▁▁▁▂▅▄▄▂▄▆█▄ beats 88/246	medium · avg 62%	▇ 0.8%	at its level
Decision Index 0.4.1: ACOS (per-review F1) submitted	19.2% ✎	▁··▂▃▃▅▇█▇▃▃ beats 104/246	hard · avg 22%	▇ 0.8%	noisy
Decision Index 0.4.1: RAGTruth (F1 on hallucinated class) submitted	52.2% ✎	▁▁▁▁▁▁▁▁▁▃▆█ beats 58/246	easy · avg 66%	▇ 0.6%	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	86.0% ✎	▁···▁▁▁▁▃▆█▃ beats 106/246	easy · avg 84%	▇ 0.6%	noisy

arts & human taste▁▇▄▃▅▅█▅▄▂▁▁ #144 of 247 · 7.4% of its score · 6 tests

Beats OneJev1.6-4B, Nace.AI Drex v1.5, Jegue 0.1 (9B) 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	62.9% ✎	▁▁▂▃▃▃▇▅█▅▃▁ beats 120/246	medium · avg 61%	▇ 3.2%	at its level
Decision Index 0.4.1: Humicroedit (accuracy) submitted	59.7% ✎	▁·▁▃▂▂▅█▄▂▁▁ beats 88/246	medium · avg 60%	▇ 1.5%	blunt · at its level
Decision Index 0.4.1: BPoMP (accuracy) submitted	83.4% ✎	▁█▄▄▅█▆▅▄▁▁▁ beats 103/246	easy · avg 83%	▇ 1.1%	sharp · noisy
Decision Index 0.4.1: cfcolor (accuracy) submitted	60.5% ✎	▁········▁▇█ beats 106/246	medium · avg 60%	▇ 0.8%	blunt · at its level
Decision Index 0.4.1: POP909-CL (accuracy) submitted	10.0% ✎	▄▅▅▃▆▅▃▅▂▅▃█ beats 103/237	hard · avg 21%	▇ 0.5%	sharp · noisy
Decision Index 0.4.1: Habermas Machine (accuracy) submitted	40.6% ✎	▁▂▄█▇▅▁▁▄▄▁▁ beats 77/246	medium · avg 47%	▇ 0.2%	blunt · at its level

knowledge & reasoning▁▂▂▂▂▃▃▅▅▆█▂ #142 of 263 · 43% of its score · 19 tests

Beats Winnow-12B (DecisionBench b6ac22b0), Amber Hybrid 4B (decision readout), StartLux-Decision v1.1 2B here, though they rank higher overall. Trails razorback16 diffgemma (decision readout), Solar Decide Flash, mmastrac diffgemma (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	51.6% ✎	▄▄▃▃▂▄▅▇██▅▁ beats 119/246	medium · avg 46%	▇▇▇ 7.9%	sharp · at its level
Decision Index 0.4.1: MMLU (accuracy) submitted	74.2% ✎	▁▄▄▃▃▂▅▇▅█▇▁ beats 116/246	easy · avg 70%	▇▇ 6.1%	sharp · at its level
Decision Index 0.4.1: BBH (accuracy) submitted	60.9% ✎	▄▆▄▄█▇▇▄▃▁▁▁ beats 85/246	easy · avg 66%	▇▇ 5.3%	at its level
Decision Index 0.4.1: ARC-Challenge (accuracy) submitted	93.9% ✎	▂▂▂▂▂▂▁▃▄▆█▆ beats 124/246	very easy · avg 90%	▇▇ 4.3%	sharp
Decision Index 0.4.1: WinoGrande (accuracy) submitted	80.7% ✎	█▅█▅▄█▆▄▃▃▁▁ beats 129/246	easy · avg 78%	▇ 2.7%	typical
Decision Index 0.4.1: GSM8K (accuracy (4- and 10-choice)) submitted	40.6% ✎	▁··▄█▆▆▆▅▅▃▁ beats 112/246	medium · avg 45%	▇ 2.6%	sharp · at its level
Decision Index 0.4.1: ARC-Easy (accuracy) submitted	97.8% ✎	▁▁▂▁▂▂▂▂▃▆█▅ beats 119/246	very easy · avg 96%	▇ 2.4%	sharp · maxed out · noisy
Decision Index 0.4.1: GPQA Diamond (accuracy) submitted	41.3% ✎	▂▆▅▅▆█▆▃▁··▁ beats 122/246	medium · avg 39%	▇ 2.3%	blunt · at its level

11 more tests

test	result	vs the others in its run	difficulty	counts	why
Ollaya · PubMedQA	74.8% ✎	▃▂▃▃·▂▃▃▆█▆▂ beats 20/25	easy · avg 69%	▇ 1.8%	at its level
Ollaya · BoolQ	86.0% ✎	▁··▃▃▁▄▁▃█▇▄ beats 17/25	easy · avg 84%	▇ 1.4%	typical
Decision Index 0.4.1: MuSR (accuracy) submitted	53.7% ✎	▁▁▂█▅▄▁▁▁▁▁▁ beats 51/246	medium · avg 60%	▇ 0.9%	blunt · at its level
Decision Index 0.4.1: SATA-Bench (case exact accuracy) submitted	31.6% ✎	▁▁▁▁▁▁▁▂▅█▂▁ beats 165/246	hard · avg 22%	▇ 0.9%	typical
Decision Index 0.4.1: CRUXEval (accuracy) submitted	43.7% ✎	██▄▃▃▄▁▁···▁ beats 79/246	medium · avg 58%	▇ 0.9%	at its level · noisy
Decision Index 0.4.1: CLadder (accuracy) submitted	63.3% ✎	▁███▆▅▃▃▁▁▁▁ beats 80/246	easy · avg 71%	▇ 0.8%	at its level
Ollaya · SQuAD 2 answerability	80.9% ✎	▁▁▃▁▇█▃▃▄▃·▁ beats 16/25	easy · avg 80%	▇ 0.7%	blunt
Ollaya · VitaminC verification	79.1% ✎	▁···▃·▄█▃··▁ beats 19/25	easy · avg 78%	▇ 0.7%	blunt
Decision Index 0.4.1: ChessBench (accuracy) submitted	10.5% ✎	▄▇▄▄▄▆▄█▄▁▁▁ beats 65/246	hard · avg 16%	▇ 0.5%	blunt
Ollaya · MultiNLI	91.0% ✎	▂··▄██▂▆▆█▆▂ beats 21/25	very easy · avg 87%	▇ 0.4%	blunt
Decision Index 0.4.1: HLE (accuracy) submitted	10.4% ✎	▁▁▂▃▆█▇▅▅▃▁▁ beats 44/246	very hard · avg 13%	▇ 0.0%	blunt

workflows▁▁▁▂▁▂▂▆█▄▁▁ #55 of 114 · 11% of its score · 9 tests

Beats Clef-omni, Clef-flash, Bespoke Nimble 9B v3 here, though they rank higher overall. Trails StartLux-Decision v1.1 0.8B, JevK5 v0.3, Together Tev1 Qwen3.5-4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Ollaya · Typed decisions test	67.8% ✎	▄·▂▆▄▄▂·▄█▆▆ beats 15/22	medium · avg 64%	▇ 2.9%	at its level
Ollaya · PAWS paraphrase	79.2% ✎	▂▂▂▂·▃█▆·▂·▁ beats 10/25	easy · avg 81%	▇ 1.7%	typical
Ollaya · SummEval relevance	45.8% ✎	▁··▁▂▁▅··▃██ beats 15/25	medium · avg 41%	▇ 1.5%	at its level
Ollaya · MASSIVE German intent	86.0% ✎	▁····▁·▂▃▆█▃ beats 19/23	easy · avg 84%	▇ 1.5%	sharp
Ollaya · Aegis 2 safety	80.4% ✎	▂·▁▁▂·▃▃▂█▃▁ beats 17/25	easy · avg 77%	▇ 1.4%	typical
Ollaya · SummEval consistency	79.2% ✎	▁·▁▁··▁▁▁▁▄█ beats 13/25	easy · avg 75%	▇ 0.9%	sharp · noisy
Ollaya · MASSIVE English intent	86.6% ✎	▁·······▂▃█▃ beats 18/23	very easy · avg 86%	▇ 0.8%	noisy
Ollaya · HelpSteer 2 scoring	45.4% ✎	▁▁··▁·▁▃▇█▅▄ beats 22/23	medium · avg 38%	▇ 0.6%	blunt · at its level

1 more test

test	result	vs the others in its run	difficulty	counts	why
Ollaya · Civil Comments	85.3% ✎	▂▂▂·▂▅▅▆▃▅·█ beats 10/23	very easy · avg 86%	▇ 0.1%	blunt · noisy

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)	36.5700 points ✎	▂▂▂▃▃▃▄▅▅▇█▃ beats 85/247	avg 41.206 points	▇▇ 4.0%	blunt
Decision Index 0.4.1: private new domains (equated points)	35.9100 points ✎	▃█▃▃▂▆▃▃▄▄▄▄ beats 107/247	avg 38.861 points	▇ 2.2%	blunt

$ where to run it

Other ways to run it

Run it as published · needs 11.1 GB, NVIDIA only · from github.com/ollaya-dev
ollaya run decider:4b --preset triage "The customer requests a refund."

Download model files ↗
Published model files verified. This is a verified repository snapshot; the benchmark does not establish an identical weight revision. A matching Ollaya runtime manifest and published short-workload GPU memory measurement were reviewed separately; see local variants and identity_note. The command was not executed in this review. Source · checked 2026-10-31.

Model repository and documentation: huggingface.co/Mapika ↗

$ how the score was worked out · Decision Index 0.4.1 82%, Ollaya · original runs 16%, Citead GEO Citation Benchmark… 1%

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	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 82%	CC BY 4.0
Ollaya · original runs	14	▇▇▇▇▇ 16%	Apache-2.0
Citead GEO Citation Benchmark (October 2026)	1	▇ 1%	CC BY-SA 4.0
prior · pull toward the middle		▇ 1%
weighted average		= 50.6

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]	20.0%	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