Best Decision Models, Ranked

A decision model answers typed questions (choose one option, score on a rubric, yes or no) with a probability for each allowed answer instead of generating text. 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.

Ranked list (top 25)

#	Model	Decision Index	ECE	p50 latency	Per 1k decisions	Params
1	Fastino GLiDE Fastino	71.5	0.020	93 ms	$0.030	28B
2	Lychee-27B LycheeAIOpen weights	67.3	0.108	97 ms	—	26B
3	vLLM-SR d3 (27B) vLLM Semantic RouterOpen weights	67.2	0.043	97 ms	—	26B
4	linne-oak-1-0 MISA-AICOREOpen weights	66.9	0.105	105 ms	—	31B
5	Decider 31B (NVFP4) MapikaOpen weights	66.0	0.024	72 ms	—	33B
6	bobcat-1.3 sanghwa-naOpen weights	64.8	0.102	64 ms	—	31B
7	amos-1-large Scira (mukaddamzaid)Open weights	64.8	0.032	148 ms	—	—
8	frontier-infra Jebadiah 27B v2.1 Frontier InfraOpen weights	64.4	0.056	303 ms	—	28B
9	Sieve-31B Sthānika AIOpen weights	64.3	0.033	253 ms	—	31B
10	AnyJev L0 · Gemma-4-31B morriszjmOpen weights	63.6	0.240	98 ms	—	31B
11	deck31b (Gemma 4 31B) krishna-gogineni-765Open weights	63.0	0.204	219 ms	—	33B
12	Darwin-27B-ZTC FINAL_BenchOpen weights	63.0	0.045	104 ms	—	26B
13	KnowLine-35B-A3B-Gen1 PelaAIOpen weights	62.9	0.048	41 ms	—	36B
14	Perplexity Decider v1.1 (27B) PerplexityOpen weights	62.8	0.071	104 ms	$0.0040	28B
15	decisio 0.6.0 (Gemma 4 31B) aminryOpen weights	62.7	0.142	54 ms	—	31B
16	Decider chat · Gemma-4-31B MapikaOpen weights	61.6	0.047	109 ms	—	33B
17	Torchcast Decision 27B Torchcast AIOpen weights	61.5	0.024	100 ms	—	28B
18	Qwen Decision 27B noncausal LoRA jsaurabhOpen weights	61.1	0.051	127 ms	—	28B
19	Quyet-1.0-Large chinhncOpen weights	60.8	0.078	105 ms	—	33B
20	Jev TypeSafe AI	60.2	0.080	196 ms	$0.0084	—
21	Sieve 27B Sthānika AIOpen weights	60.2	0.065	101 ms	—	28B
22	OneJev1.5-27B OmniJevOpen weights	59.9	0.024	224 ms	—	27B
23	mogan-decision-31B-it Mogan AIOpen weights	59.8	0.054	109 ms	—	31B
24	Shisa DE-2 (repeat) Shisa.AIOpen weights	59.8	0.207	57 ms	—	26B
25	linne-fern 1.0 (27B) MISA-AICOREOpen weights	59.6	0.079	108 ms	—	28B

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 calibrated Cheapest Open weights Runs on CPU Jev alternatives

Frequently asked questions

What is the best decision model?

On the Jev Decision Index, Fastino GLiDE ranks first with a Full score of 71.5.

How is a decision model different from an LLM?

An LLM generates text. A decision model returns a probability distribution over the allowed answers for each typed question, so its output can be thresholded, audited and calibrated without parsing text.

Is the most accurate decision model the best choice?

Not always. Calibration (how far stated confidence is from the real hit rate), latency and cost per 1,000 decisions often matter as much; see the calibration and cost rankings.