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

Best Decision models for images & screens to run locally on a 16GB graphics card

A leaderboard of the best Decision models for images & screens you can run locally on a 16GB graphics card (with 32GB of system RAM). Clef-flash is #1 right now, followed by Perplexity Decider Flash v1.2 (4B) and vLLM-SR d3-flash (9B). 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.

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 → ~16GB fast · 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]

vram[4][8][12][16][24][32][48][64][80][96][128]GB

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 vision evaluations only. Compare scores within this task. Models without task measurements are excluded.

New #1s for your hardware:

orRSS

detecting your OS…

$ terminal

Score vs memory to run

     │                                                          ●─Perplexity Decider Flash …───Clef-flash─●
  50 ┤ · · · · · · · · · · · · ··· · · · · · · · · · · · · · · ·│· ·∘· · · · · · · · · · · ∘ · ··· · · · · · ·
     │   ●────────────────────────────────────────────●─vLLM-SR d3-nano (2B)
  25 ┤ · · · · · · · · · · · · ·∘· · · · · · · · · · · · · · · ··· · · · · · · · · · · · · · · ··· · · · · · ·
     │   ∘
     └──────────────────────────┬───────────────────────────────┬───────────────────────────────┬─────────────
                               5GB                            10GB                            20GB

#	model	task score↓	where↓	RAM↓	speed↓	get it	sources

#	model	task score↓	where↓	RAM↓	get it
14	Clef-flash · bf16 Cloudflare	64±3	local (slow)	~25GB	how to run ›
21	Perplexity Decider Flash v1.2 (4B) · ollaya-f32 Perplexity	62±3	local	~10GB	how to run ›
35	vLLM-SR d3-flash (9B) · ollaya-f32 vLLM Semantic Router	55±5	local (slow)	~18GB	how to run ›
40	vLLM-SR d3-mini (4B) · ollaya-f32 vLLM Semantic Router	51±5	local	~11GB	how to run ›
62	vLLM-SR d3-nano (2B) · ollaya-f32 vLLM Semantic Router	39±5	local	~8.2GB	how to run ›
64	LiquidAI d1-omni-600M · ollaya-f32 Liquid AI	36±14*	local	~2.7GB	how to run ›
75	vLLM-SR d3-lite (0.8B) · ollaya-f32 vLLM Semantic Router	24±5	local	~4.5GB	how to run ›
79	vLLM-SR d3-edge (0.6B) · ollaya-f32 vLLM Semantic Router	10±5	local	~2.6GB	how to run ›

+ 71 more don't fit your machine or budget · show them

79 of 79 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