Best Decision Models That Run on a CPU

CPU-class decision models (encoders and small decoders, under about 1.5B parameters) run next to the data with no GPU. Models are ranked by their Full score on the Jev Decision Index, a third-party board of 37 benchmarks with private held-out sets, until scores from our own Metatext Decision Benchmark are published. Latency is the board's median per call on its GPU runner.

Ranked list (top 25)

#	Model	Decision Index	ECE	p50 latency	Per 1k decisions	Params
1	OneJev1.5-0.8B OmniJevOpen weights	29.0	0.035	83 ms	—	1B
2	vLLM-SR d3-lite (0.8B) vLLM Semantic RouterOpen weights	26.2	0.094	37 ms	—	853M
3	jiwo 0.8B v2 eljiwoOpen weights	25.8	0.068	36 ms	—	752M
4	Saina Helm 2.0 SainaOpen weights	25.3	0.168	39 ms	—	853M
5	crivo-400m tardellirsOpen weights	23.7	0.121	17 ms	—	396M
6	Sieve 0.8B Sthānika AIOpen weights	20.8	0.088	52 ms	—	590M
7	EXAONE-4.0-1.2B-JEV v0.3 carrtesyOpen weights	20.6	0.079	149 ms	—	1B
8	tasksource Decider nano v4 (150M) tasksourceOpen weights	19.9	0.070	13 ms	—	149M
9	vLLM-SR Decision 2.0 Eos 0.8B vLLM Semantic RouterOpen weights	18.4	0.130	25 ms	—	870M
10	Crivo-150M tardellirsOpen weights	18.3	0.115	13 ms	—	150M
11	Bekko System One v0 400M hotchpotchOpen weights	18.0	0.111	17 ms	—	400M
12	vLLM-SR d3-edge (0.6B) vLLM Semantic RouterOpen weights	17.7	0.141	19 ms	—	594M
13	DecisionTune 1.0 DecisionTuneOpen weights	17.2	0.147	14 ms	—	395M
14	Valen-0.8B Valen-TeamOpen weights	16.6	0.108	128 ms	—	854M
15	vLLM-SR Decision 2.0 Kai 0.6B vLLM Semantic RouterOpen weights	16.4	0.038	19 ms	—	600M
16	Kev 0.8B r15 Jared PalmerOpen weights	16.3	0.040	45 ms	—	870M
17	JPT-0.8B kirpOpen weights	15.5	0.069	118 ms	—	873M
18	jev-ettin-400m-20261013-v0 ElfsongOpen weights	15.2	0.090	16 ms	—	400M
19	Sifr 0.8B v3.1 mohamedlotfy50Open weights	14.5	0.120	35 ms	—	870M
20	Crivo-32M tardellirsOpen weights	14.2	0.133	8 ms	—	32M
21	Dinah-0 LukitaduarteOpen weights	13.9	0.104	14 ms	—	150M
22	Intern-Decision-0.8B Intern Large ModelsOpen weights	12.7	0.025	34 ms	—	873M
23	Jev-Style 0.8B v3 chaoliangUNSWOpen weights	12.5	0.084	36 ms	—	870M
24	Tev1-0.8B-experimental TogetherOpen weights	12.5	0.126	29 ms	—	873M
25	Bekko System One v0 68M hotchpotchOpen weights	12.4	0.113	12 ms	—	68M

Scores, calibration (ECE) and latency from the Jev Decision Index; cost per 1,000 decisions at 200 input tokens each. Updated daily.

More decision-model rankings

Best decision model Best calibrated Cheapest Open weights Jev alternatives

Frequently asked questions

What is the best decision model that runs on a CPU?

OneJev1.5-0.8B ranks first among CPU-class decision models on the Jev Decision Index.

How accurate are CPU decision models compared with large ones?

They trail the 27B-class leaders by a wide margin on the Decision Index, but answer in milliseconds on commodity hardware, which suits high-volume routing and filtering.