whatmodel.sh · find the best decision models you can run

Best Decision models for document & agent workflows to run locally

A leaderboard of the best Decision models for document & agent workflows you can run on your own computer. Clef is #1 right now, followed by Perplexity Decider Flash v1.2 (4B) and Winnow-12B (DecisionBench b6ac22b0). Published typed-decision evaluations, with evaluator-specific protocols and tested settings retained. Each task score uses only its reviewed benchmarks. Each model shows the download that fits and the command to run it.

28 models · 108 benchmarks · updated

Compare tested decision models and inference recipes across text, vision and workflow tasks. Use task filters and raw results: probability consistency measures coherence, not correctness.

Filter by your hardware and budget · local only · reset
ram[2][4][8][16][24][32][48][64][96][128][192][256]
[]
GB
gpu[apple silicon][graphics card][unified memory][none]
price[local][$0.05][$0.1][any]max per M input tokens
Filter by task
task[all tasks][tools & automation][retrieval & classification][language understanding][arts & human taste][knowledge & reasoning][vision][workflows][consistency]

Scores use reviewed workflows evaluations only. Compare scores within this task. Models without task measurements are excluded.

New #1s:
orRSS

detecting your OS…

$ terminal

detecting your OS…

Run curl "whatmodel.sh/decision?max=0"to get this list in your terminal (add &format=json for agents).

Score vs release date

Release dates available for 9 of 28 models. Undated models are omitted; month-only dates are plotted at the start of the month.

#modeltask score↓where↓RAM↓speed↓get itsources
#modeltask score↓where↓RAM↓speed↓get itsources
24Clef Cloudflare61±1local~68GBhow to run › DecisionPivot · StartLux LabsLeanMCP · JevBenchSystem One Bench (Ali Alhashimi)DecisionModelHub · support benchmarksDecideBench (Cho Yin Yong)
27Perplexity Decider Flash v1.2 (4B) Perplexity61±2local~10GBhow to run › Ollaya · original runsSystem One Bench (Ali Alhashimi)
39Winnow-12B (DecisionBench b6ac22b0) EldanRing58±1local~16GBhow to run › Hanno Labs DecisionBench v1 reviewed resultsOllaya · original runs
41vLLM-SR d3-flash (9B) vLLM Semantic Router58±2local~18GBhow to run › Ollaya · original runs
45vLLM-SR d3-mini (4B) vLLM Semantic Router57±2local~11GBhow to run › Ollaya · original runs
46Jebadiah 27B v2.1 (Ollaya Q4_K_M) frontier-infra57±4local~16GBhow to run › Ollaya · original runs
48Jebadiah 9B v2.1 (Ollaya Q8_0) frontier-infra57±4local~9.1GBhow to run › Ollaya · original runs
57Decider 4B Mapika55±2local~11GBhow to run › Ollaya · original runs
59vLLM-SR Decision 2.0 Nox 4B vLLM Semantic Router54±2local~12GBhow to run › Ollaya · original runs
61vLLM-SR Decision 2.0 Lux 9B vLLM Semantic Router54±2local~20GBhow to run › Ollaya · original runs
65Jebadiah 4B v2 (Ollaya Q8_0) frontier-infra54±2local~5.5GBhow to run › Ollaya · original runs
66Kev 4B r10 Jared Palmer54±1local~11GBhow to run › Ollaya · original runsJevalKit (Valen Team)System One Bench (Ali Alhashimi)
68Clef-flash Cloudflare53±1local~25GBhow to run › DecisionPivot · StartLux LabsLeanMCP · JevBenchOllaya · original runsJevalKit (Valen Team)DecisionModelHub · support benchmarksSystem One Bench (Ali Alhashimi)+1
69Bespoke Nimble 9B v3 Bespoke Labs52±2local~21GBhow to run › Ollaya · original runs
71Bespoke Nimble 9B v2 Bespoke Labs52±2local~20GBhow to run › Ollaya · original runs
73Clef-omni Cloudflare52±2local~87GBhow to run › LeanMCP · JevBenchSystem One Bench (Ali Alhashimi)
75vLLM-SR d3-nano (2B) vLLM Semantic Router52±2local~8.2GBhow to run › Ollaya · original runsSystem One Bench (Ali Alhashimi)
79GLiClass v4 edge (0.4B) Knowledgator Engineering49±2local~2.7GBhow to run › Ollaya · original runs
82vLLM-SR Decision 2.0 Sol 2B vLLM Semantic Router46±2local~11GBhow to run › Ollaya · original runs
83vLLM-SR d3-lite (0.8B) vLLM Semantic Router45±2local~4.5GBhow to run › Ollaya · original runsSystem One Bench (Ali Alhashimi)
84LiquidAI d1-omni-600M Liquid AI45±2local~2.7GBhow to run › Ollaya · original runsSystem One Bench (Ali Alhashimi)
86Decision-1.0-Eos-0.8B (DecisionBench 3c2d6326) llm-semantic-router44±1local~5.8GBhow to run › Hanno Labs DecisionBench v1 reviewed resultsOllaya · original runs
87vLLM-SR Decision 2.0 Eos 0.8B vLLM Semantic Router44±2local~5.8GBhow to run › Ollaya · original runs
88vLLM-SR Decision 2.0 Kai 0.6B vLLM Semantic Router43±2local~3.7GBhow to run › Ollaya · original runs
90Kev 0.8B r15 Jared Palmer41±2local~4.6GBhow to run › Ollaya · original runsSystem One Bench (Ali Alhashimi)
92vLLM-SR d3-edge (0.6B) vLLM Semantic Router40±2local~2.6GBhow to run › Ollaya · original runsSystem One Bench (Ali Alhashimi)
104laya (DecisionBench 1c5edc17) ConvAI Innovations27±1local~1.7GBhow to run › Hanno Labs DecisionBench v1 reviewed resultsOllaya · original runs
113CLM-v0.1-8B Qwen3-8B heads Contrastive-LM15±2local~15GBhow to run › Ollaya · original runsDecideBench (Cho Yin Yong)Perch benchmark results

+ 89 more don't fit your machine or budget · show them

117 of 117 catalogue models listed here · sources and licences

* new or thinly tested: ranked on too few benchmark results · † no independently run results verified · ~RAM is estimated; full context may need more memory

28 models · 108 benchmarks · updated