Kev 0.8B r15

+ contribute

Kev 0.8B r15 is an open-weights decision model from Jared Palmer, released 24 September 2026. Its published size is 0.87B parameters. Licence: apache-2.0.

Catalogue identity notes

Exact Decision Index tested engine kev-0.8b-r15; reported model Kev 0.8B r15. Type: LoRA + head; technique: autoregressive. 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: kev:0.8b, manifest 8a0d4094df504ca7329bc999dabbb89c15ad5221f55728acde0ff61c2b3909f2, matching the independent accuracy-run manifest. Native family implementation; published runtime precision label F32. Memory is measured peak GPU increase (4650 MiB, rounded upward to 4.6 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

Release date ↗ · Weights licence ↗ · Parameter count ↗

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

30.0±0.9#306 of 355 decision modelsfrom 72 benchmark results across 3 sources

$ how it compares

The decision models ranked either side of Kev 0.8B r15, out of 357 with a score, then more from Jared Palmer.

#	model	score	where	RAM	released
306	Together Tev1 Qwen3.5-0.8B Together	31±1	model files
307	Dinah-1 Lukitaduarte	30±1	model files
308	Kev 0.8B r15 Jared Palmer	30±1	local	~4.6GB	2026-09-24
309	PICK-150M v5 lCREEl	30±1	model files		2026-10-18
310	InternLM Intern-Decision Qwen3.5-0.8B Shanghai AI Laboratory (InternLM)	30±1	model files
	more from Jared Palmer
88	Kev 27B Jared Palmer	62±1	model files		2026-09
159	kev-4b (DecisionBench 485ace87) Jared Palmer	55±2	model files
162	kev-8b (DecisionBench c80773da) Jared Palmer	55±2	model files

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

$ what it's good at

tools & automation▁▃▁▁▃▂▃▄▆██▆ #216 of 250 · 7.0% of its score · 6 tests

Beats Cam1, Valen-0.8B, DM-JEPA 1.1 here, though they rank higher overall. Trails Prumo (rule-based engine), JevK5 Lite, Crivo-32M v2 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	70.5% ✎	▁▂▁▁▃▂▂▄▅▆█▁ beats 33/246	very easy · avg 92%	▇▇ 4.2%	sharp · at its level · noisy
Decision Index 0.4.1: API-Bank (accuracy) submitted	43.7% ✎	▁▁▁▁▁▁▁▃▃▂█▂ beats 37/236	easy · avg 70%	▇ 1.4%	sharp · at its level · noisy
Decision Index 0.4.1: When2Call (accuracy) submitted	43.2% ✎	▁▁▁▂▃▅▅▆█▇█▃ beats 31/246	easy · avg 72%	▇ 0.7%	at its level · noisy
Decision Index 0.4.1: ToolRet (nDCG@10) submitted	51.7% ✎	▁▁▁▁▁·▁▁▃▄█▁ beats 33/246	medium · avg 60%	▇ 0.5%	blunt · at its level
Decision Index 0.4.1: Home appliance simulator (case exact accuracy) submitted	0.0% ✎	█▁▁▃▄▄▆▄▁▂▁▃ beats 35/243	medium · avg 37%	▇ 0.1%	sharp · too hard for it · noisy
Decision Index 0.4.1: RouterBench (selected quality (quality objective)) submitted	73.1% ✎	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▁▁▁·▁▃▁▁▃▃█▅ #241 of 258 · 11% of its score · 13 tests

Beats DM-JEPA 1.1, Together Tev1 Qwen3.5-0.8B here, though they rank higher overall. Trails Prumo (rule-based engine), MoganAI Lavoir ModernBERT-large, Bekko System One v0 68M here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
System One Bench: MASSIVE intent (Arabic)	51.5% ✎	▄·▁·▁▃▃▃█▄▇▇ beats 0/29	easy · avg 76%	▇ 2.8%	at its level
System One Bench: MIRACL pairwise (Arabic)	67.0% ✎	▃···▁··▃▄▂█▇ beats 1/29	easy · avg 84%	▇ 2.4%	at its level
System One Bench: MIRACL rerank (Arabic)	62.5% ✎	▁▃·▁▁·▂·▄▄█▇ beats 2/29	easy · avg 79%	▇ 1.8%	at its level
Decision Index 0.4.1: Amazon ESCI (macro-F1) submitted	22.9% ✎	▁▁▁▁▁▃▂▂▃▃█▅ beats 14/246	medium · avg 47%	▇ 0.9%	at its level
Decision Index 0.4.1: BRIGHT (nDCG@10) submitted	31.8% ✎	▁▁▁▁▁▁▁▂▃▄▇█ beats 57/246	medium · avg 38%	▇ 0.9%	blunt · at its level
System One Bench: MASSIVE scenario (Arabic)	38.5% ✎	▃▁·▄▃▂█▂▁▁·▃ beats 1/29	medium · avg 60%	▇ 0.7%	at its level
Decision Index 0.4.1: CLINC150+OOS (macro-F1) submitted	15.0% ✎	▁▁▁▁▃▁▁▂▃▆█▄ beats 14/238	easy · avg 83%	▇ 0.4%	sharp · at its level · noisy
Decision Index 0.4.1: HoVer (accuracy) submitted	51.0% ✎	▆▆▇▅▄▄▃▇▄█▆▄ beats 12/246	easy · avg 74%	▇ 0.4%	blunt · at its level

5 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: PhishNChips phishing decisions (accuracy) submitted	51.6% ✎	█▅▅▄▃▂▁▁▁▁·▁ beats 54/245	easy · avg 70%	▇ 0.3%	at its level · noisy
Decision Index 0.4.1: BANKING77 (macro-F1) submitted	77.6% ✎	▁·▁▁▁▁▁▁▂█▇█ beats 73/238	easy · avg 79%	▇ 0.3%	at its level · noisy
System One Bench: AG News (English)	91.0% ✎	▁···▅▁▅▂▆█▃▅ beats 25/29	easy · avg 88%	▇ 0.2%	blunt
System One Bench: Banking77 (English)	80.5% ✎	█▅▆█▅▃▃▃···▂ beats 3/26	very easy · avg 87%	▇ 0.0%	blunt · noisy
Decision Index 0.4.1: SGD/SGD-X (macro-F1) submitted	53.6% ✎	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▂▃▂▃▂▅▆▃█▆▄▃ #233 of 256 · 23% of its score · 18 tests

Trails JevK5 Lite, LFM2.5-2.6B-RLCD, JEV Ettin 400M 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	42.9% ✎	▂▄▂▂▄▇▃▇█▅▃▁ beats 23/246	very easy · avg 85%	▇▇▇ 6.8%	sharp · at its level
Decision Index 0.4.1: NLI4CT (macro-F1) submitted	55.9% ✎	▁▁▁▂▃▃▃▄▇▇█▄ beats 28/246	easy · avg 74%	▇ 3.7%	at its level
Decision Index 0.4.1: ANLI (macro-F1) submitted	33.6% ✎	▁▂▃▅▅█▂····▁ beats 21/246	medium · avg 60%	▇ 2.7%	at its level
System One Bench: BoolQ (English)	76.0% ✎	▁▃▁▃·▁▃▃▇▇█▅ beats 1/29	easy · avg 86%	▇ 1.7%	at its level
System One Bench: OSACT4 offensive (Arabic)	49.0% ✎	▃······▁▃█▇▄ beats 0/29	easy · avg 84%	▇ 1.6%	sharp · at its level
System One Bench: SST-2 (English)	90.0% ✎	▁·▃▁▃▂▃▅█·▂▁ beats 5/29	easy · avg 92%	▇ 1.5%	at its level
Decision Index 0.4.1: iSarcasmEval (Sarcasm F1 · track A, English) submitted	19.9% ✎	▁·▁·▁▁▂▄▅█▇▄ beats 24/246	medium · avg 44%	▇ 0.8%	typical
Decision Index 0.4.1: ContractNLI (macro-F1) submitted	40.9% ✎	▁▁▁▁▁▃▁▂▂▇▇█ beats 24/246	easy · avg 71%	▇ 0.7%	at its level · noisy

10 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: FinEntity (macro-F1) submitted	71.4% ✎	▁···▁▁▁▁▃▆█▃ beats 44/246	easy · avg 84%	▇ 0.7%	at its level · noisy
Decision Index 0.4.1: VAST (macro-F1) submitted	40.6% ✎	▁▁▁▂▅▄▄▂▄▆█▄ beats 34/246	medium · avg 62%	▇ 0.5%	at its level
System One Bench: SST-5 (English)	47.0% ✎	▃·▁▁▃·▇▄▄▄█▅ beats 5/29	medium · avg 51%	▇ 0.5%	blunt · at its level
Decision Index 0.4.1: RAGTruth (F1 on hallucinated class) submitted	46.2% ✎	▁▁▁▁▁▁▁▁▃▃▆█ beats 42/246	easy · avg 66%	▇ 0.4%	at its level · noisy
System One Bench: XNLI French (French)	66.0% ✎	▂▅·▁▁▅▅█▃▁·▃ beats 4/29	medium · avg 76%	▇ 0.3%	blunt · at its level
System One Bench: XNLI Arabic (Arabic)	58.7% ✎	▄▁▃▁▄▇█▅▃▁·▃ beats 4/29	medium · avg 68%	▇ 0.3%	at its level
Decision Index 0.4.1: ACOS (per-review F1) submitted	9.9% ✎	▁··▂▃▃▅▇█▇▃▃ beats 55/246	hard · avg 22%	▇ 0.2%	noisy
System One Bench: Tweet sentiment (Arabic)	57.0% ✎	▁······▁▁▄█▄ beats 3/29	medium · avg 62%	▇ 0.2%	blunt · at its level
System One Bench: XNLI English (English)	81.3% ✎	▃·▁·▃▇█▄█▃▃▁ beats 3/29	easy · avg 85%	▇ 0.1%	blunt · at its level · noisy
System One Bench: DAIR Emotion (English)	56.0% ✎	██▃·····▁··▁ beats 8/29	medium · avg 59%	▇ 0.0%	blunt · at its level

arts & human taste▁▇▄▃▅▅█▅▄▂▁▁ #243 of 247 · 5.0% of its score · 6 tests

Beats JEV Ettin 150M, JEV FLAN-T5 Large public v2 (342M) here, though they rank higher overall. Trails Zev (decision readout), LFM2.5-350M-RLCD, Nirnaya 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	27.8% ✎	▁▃▂▃▃▃▇▅█▅▃▁ beats 14/246	medium · avg 61%	▇ 2.0%	at its level
Decision Index 0.4.1: BPoMP (accuracy) submitted	47.7% ✎	▁█▄▄▅█▆▅▄▁▁▁ beats 5/246	easy · avg 83%	▇ 1.2%	sharp · at its level · noisy
Decision Index 0.4.1: Humicroedit (accuracy) submitted	50.9% ✎	▁·▁▃▂▂▅█▄▂▁▁ beats 27/246	medium · avg 60%	▇ 1.0%	blunt · at its level
Decision Index 0.4.1: cfcolor (accuracy) submitted	55.4% ✎	▁········▁▇█ beats 54/246	medium · avg 60%	▇ 0.5%	blunt · at its level
Decision Index 0.4.1: Habermas Machine (accuracy) submitted	33.4% ✎	▁▃▄█▇▅▁▁▄▄▁▁ beats 10/246	medium · avg 47%	▇ 0.1%	blunt · at its level
Decision Index 0.4.1: POP909-CL (accuracy) submitted	1.5% ✎	▄▅▅▃▆▅▃▅▂▅▃█ beats 32/237	hard · avg 21%	▇ 0.1%	sharp · too hard for it · noisy

knowledge & reasoning▁▂▂▃▂▃▃▅▅▆█▂ #217 of 263 · 39% of its score · 19 tests

Beats vLLM-SR vela3-fast (Vela 3.0, 0.4B), Decision-1.0-Eos-0.8B (DecisionBench 3c2d6326), Amber 0.4B here, though they rank higher overall. Trails laya (DecisionBench 1c5edc17), LFM2.5-2.6B-RLCD here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: ARC-Easy (accuracy) submitted	82.2% ✎	▁▁▂▁▂▃▂▂▃▆█▅ beats 47/246	very easy · avg 96%	▇▇▇ 8.5%	sharp · at its level · noisy
Decision Index 0.4.1: ARC-Challenge (accuracy) submitted	63.3% ✎	▂▂▂▂▃▂▁▃▄▆█▆ beats 45/246	very easy · avg 90%	▇▇▇ 7.5%	sharp · at its level
Decision Index 0.4.1: MMLU (accuracy) submitted	45.7% ✎	▁▄▄▃▃▂▅▇▅█▇▁ beats 46/246	easy · avg 70%	▇▇ 4.5%	sharp · at its level
Decision Index 0.4.1: BBH (accuracy) submitted	41.4% ✎	▄▆▄▄█▇▇▄▃▁▁▁ beats 40/246	easy · avg 66%	▇ 3.6%	at its level
Decision Index 0.4.1: MMLU-Pro (accuracy) submitted	22.3% ✎	▄▄▃▃▂▄▅▇██▅▁ beats 42/246	medium · avg 46%	▇ 3.3%	sharp
Decision Index 0.4.1: WinoGrande (accuracy) submitted	52.8% ✎	█▅█▅▄█▆▄▃▃▁▁ beats 23/246	easy · avg 78%	▇ 2.5%	at its level
Ollaya · BoolQ	79.3% ✎	▁··▃▃▁▄▁▃█▇▄ beats 8/25	easy · avg 84%	▇ 1.3%	at its level
Ollaya · PubMedQA	44.8% ✎	▃▃▃▃·▂▃▃▆█▆▂ beats 2/25	easy · avg 69%	▇ 1.3%	at its level

11 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: GPQA Diamond (accuracy) submitted	33.2% ✎	▂▆▅▅▆█▆▃▁··▁ beats 75/246	medium · avg 39%	▇ 1.2%	blunt · at its level
Decision Index 0.4.1: GSM8K (accuracy (4- and 10-choice)) submitted	18.9% ✎	▁··▄█▆▆▆▅▅▃▁ beats 27/246	medium · avg 45%	▇ 1.0%	sharp
Decision Index 0.4.1: MuSR (accuracy) submitted	48.9% ✎	▁▁▃█▅▄▁▁▁▁▁▁ beats 26/246	medium · avg 60%	▇ 0.6%	blunt · at its level
Ollaya · SQuAD 2 answerability	66.2% ✎	▁▁▃▁▇█▃▃▄▃·▁ beats 3/25	easy · avg 80%	▇ 0.6%	blunt · at its level
Decision Index 0.4.1: CLadder (accuracy) submitted	50.5% ✎	▁███▆▅▃▃▁▁▁▁ beats 16/246	easy · avg 71%	▇ 0.6%	at its level
Ollaya · VitaminC verification	66.3% ✎	▁···▃·▄█▃··▁ beats 4/25	easy · avg 78%	▇ 0.6%	blunt · at its level
Decision Index 0.4.1: CRUXEval (accuracy) submitted	37.7% ✎	██▄▃▃▄▁▁···▁ beats 39/246	medium · avg 58%	▇ 0.5%	at its level · noisy
Ollaya · MultiNLI	79.9% ✎	▂··▄██▂▆▆█▆▂ beats 4/25	very easy · avg 87%	▇ 0.4%	blunt
Decision Index 0.4.1: SATA-Bench (case exact accuracy) submitted	9.0% ✎	▁▁▁▁▁▁▃▂▅█▂▁ beats 42/246	hard · avg 22%	▇ 0.3%	typical
Decision Index 0.4.1: ChessBench (accuracy) submitted	7.9% ✎	▄▇▄▄▄▆▄█▄▁▁▁ beats 21/246	hard · avg 16%	▇ 0.2%	blunt
Decision Index 0.4.1: HLE (accuracy) submitted	13.8% ✎	▁▁▂▃▆█▇▅▅▃▁▁ beats 168/246	very hard · avg 13%	▇ 0.0%	blunt

workflows▁▁▁▂▁▃▂▆█▄▁▁ #88 of 114 · 11% of its score · 10 tests

Beats kev-0.6b (DecisionBench dece6dba), mojev (DecisionBench 0c8695b6), smalljev-semantic-v9 (DecisionBench 8e2bac7c) here, though they rank higher overall. Trails vLLM-SR Decision 2.0 Kai 0.6B, LiquidAI d1-omni-600M, InternLM Intern-Decision Qwen3.5-0.8B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Ollaya · Typed decisions test	46.0% ✎	▄·▂▆▄▄▂·▄█▆▆ beats 3/22	medium · avg 64%	▇ 2.0%	at its level
System One Bench: Typed decisions (English)	46.3% ✎	▁▃·▁▁▁▅▄▇██▁ beats 1/29	medium · avg 63%	▇ 1.8%	blunt · at its level
Ollaya · MASSIVE German intent	66.0% ✎	▁····▁·▃▃▆█▃ beats 3/23	easy · avg 84%	▇ 1.7%	sharp · at its level
Ollaya · PAWS paraphrase	66.0% ✎	▂▂▃▂·▃█▆·▂·▁ beats 5/25	easy · avg 81%	▇ 1.7%	at its level
Ollaya · Aegis 2 safety	69.2% ✎	▂·▁▁▂·▃▃▂█▃▁ beats 7/25	easy · avg 77%	▇ 1.2%	at its level
Ollaya · MASSIVE English intent	74.9% ✎	▁·······▃▃█▃ beats 3/23	very easy · avg 86%	▇ 0.9%	at its level · noisy
Ollaya · SummEval consistency	54.9% ✎	▁·▁▁··▁▁▃▁▄█ beats 6/25	easy · avg 75%	▇ 0.7%	sharp · at its level · noisy
Ollaya · SummEval relevance	13.8% ✎	▁··▃▂▁▅··▃██ beats 1/25	medium · avg 41%	▇ 0.7%	typical

2 more tests

test	result	vs the others in its run	difficulty	counts	why
Ollaya · HelpSteer 2 scoring	28.9% ✎	▁▁··▁·▃▃▇█▅▄ beats 3/23	medium · avg 38%	▇ 0.3%	blunt · at its level
Ollaya · Civil Comments	83.7% ✎	▂▂▂·▂▅▅▆▃▅·█ beats 7/23	very easy · avg 86%	▇ 0.1%	blunt · noisy

other tests2 not in a task · 4.0% of its score

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: private same skills (equated points)	14.9700 points ✎	▂▂▃▃▃▃▄▅▅▇█▃ beats 26/247	avg 41.206 points	▇ 2.6%	blunt
Decision Index 0.4.1: private new domains (equated points)	18.2900 points ✎	▃█▃▃▂▆▃▃▄▄▄▄ beats 70/247	avg 38.861 points	▇ 1.4%	blunt

$ where to run it

Other ways to run it

Run it as published · needs 4.6 GB, NVIDIA only · from github.com/ollaya-dev
ollaya run kev:0.8b --preset triage "The customer requests a refund."

Download model files ↗
Published model files verified. Includes adapters or readout weights; follow the model documentation for the required base model. 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/jaredpalmer ↗

$ how the score was worked out · Decision Index 0.4.1 70%, System One Bench (Ali Alhashi… 16%, Ollaya · original runs 13%

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	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 70%	CC BY 4.0
System One Bench (Ali Alhashimi)	16	▇▇▇▇▇ 16%	CC-BY-4.0
Ollaya · original runs	14	▇▇▇▇ 13%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 30.0

not counted (2)

Decision Index 0.4.1	Decision Index 0.4.1: RouterBench (selected quality (quality objective)) [correct]	73.1%	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]	53.6%	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