General

Decision Index Language leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

As of 2026-10-29, the highest measured score on Decision Index Language is 67.1% by mammoth-1. 6 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	mammoth-1	67.1%	measured
2	Perplexity Decider v1.2 27B	65.5%	measured
3	d1-ultra	62.6%	estimated ± 4.2 pp, medium confidence
4	Perplexity Decider Flash v1.2 4B	62.2%	measured
5	Celeris-1 Decision	62.2%	estimated ± 4.2 pp, medium confidence
6	elephant-1	62.1%	measured
7	Blink v0.3 26B-A4B (NVFP4)	59.3%	measured
8	Surogate Rune 26B-A4B v3	58.6%	measured
9	Quyet-1.0-Large	58.1%	measured
10	Jev 1.14	58.0%	measured
11	Kev 27B	57.5%	measured
12	deck31b (Gemma 4 31B)	57.4%	measured
13	Torchcast Decision 27B	57.3%	measured
14	JPT-35B-A3B	57.2%	measured
15	JADE	56.5%	measured
16	simple-jev · Qwen3.8-27B	56.1%	measured
17	GEV-26B-Decide	55.9%	measured
18	Decider chat · Gemma-4-31B	55.5%	measured
19	Eikos-27B	55.1%	measured
20	Bespoke Nimble 9B v3	54.6%	measured
21	Solomon v1.1	54.2%	measured
22	vLLM-SR Decision 2.0 Vega 27B	53.9%	measured
23	Xor 1.2	53.9%	measured
24	JevK5 9B v0.3.3	53.7%	measured
25	d1	53.6%	estimated ± 4.2 pp, medium confidence
26	Drex v2.0	52.7%	measured
27	Decider chat · Qwen3.6-27B	52.6%	measured
28	JoshuaSP diffusiongemma (open-jev)	51.9%	measured
29	JEV-27B	51.2%	measured
30	Winnow-12B Q8	50.6%	measured
31	decisio 0.4.0 (Qwen3.6-35B-A3B)	50.2%	measured
32	djev	48.6%	measured
33	decider-35b-a3b	48.4%	measured
34	Jev-Omni	48.1%	measured
35	Clef	47.8%	measured
36	mmastrac diffusiongemma (vLLM PR 57250)	47.7%	measured
37	ezjev 4B s2	47.1%	measured
38	Hopper 12B (trained)	47.1%	measured
39	Microsoft-Decision-1	46.8%	estimated ± 4.2 pp, medium confidence
40	JPT-9B	46.5%	measured
41	JevK5 v0.3	46.3%	measured
42	razorback16 openjev diffusiongemma (NVFP4, vLLM)	46.3%	measured
43	vLLM-SR Decision 2.0 Lux 9B	46.1%	measured
44	Sieve 9B	45.3%	measured
45	Jevfire	44.7%	measured
46	Bespoke Nimble 9B v2	44.4%	measured
47	Kev 9B v2	44.3%	measured
48	decisio 0.4.0 (Gemma 4 12B)	44.0%	measured
49	GLiDE	43.2%	estimated ± 4.2 pp, medium confidence
50	RSI-Jev v5.0 VL 3B	42.3%	measured
51	vLLM-SR Decision 2.0 Nox 4B	42.2%	measured
52	Lev1 3.7B	41.8%	measured
53	Intern-Decision-4B	41.6%	measured
54	Tev1-4B-experimental	41.4%	measured
55	JPT-4B	41.3%	measured
56	Mercury Decide	41.1%	estimated ± 4.2 pp, medium confidence
57	Drex v1.5	40.8%	measured
58	Clef-Omni	40.7%	measured
59	JEV-9B	40.7%	measured
60	Clef-Flash	40.5%	measured
61	Hopper (G) 1.2	39.8%	measured
62	Kas 4B	37.6%	measured
63	Winnow-E4B	37.5%	measured
64	Jevstral 8B	37.3%	measured
65	metask-jev-4b	37.0%	measured
66	open-jev (pngwn)	36.9%	measured
67	Intelif Qwen3-4B	36.6%	measured
68	Jobe	36.0%	measured
69	SemIf	35.9%	measured
70	Kev 4B r10	35.6%	measured
71	NeoHorse-Jev-4B	34.2%	measured
72	Decider 4B	33.9%	measured
73	Ryotide Qwen v0.3.1	33.2%	measured
74	openvons	31.7%	measured
75	lev	29.5%	measured
76	mini-jev	29.1%	measured
77	Intern-Decision-2B	28.8%	measured
78	Gevva E4B	28.3%	measured
79	decider-2b	27.4%	measured
80	vLLM-SR Decision 2.0 Sol 2B	24.0%	measured
81	this-that 1.2	23.4%	measured
82	EXAONE-4.0-1.2B-JEV v0.3	22.2%	measured
83	Sieve 2B	21.7%	measured
84	Gevva E2B	21.2%	measured
85	vLLM-SR Decision 2.0 Eos 0.8B	19.7%	measured
86	Bekko System One v0 400M	19.6%	measured
87	JPT-0.8B	19.3%	measured
88	GLiNER2.5-Decide	15.5%	measured
89	Bosun v3.1 1.7B	14.7%	measured
90	Von 1.3	14.7%	measured
91	Laya	14.6%	measured
92	Tev1-0.8B-experimental	14.4%	measured
93	Sifr 0.8B v3.1	14.3%	measured
94	JevK5 Lite	13.8%	measured
95	vLLM-SR Decision 2.0 Kai 0.6B	13.1%	measured
96	Kev 0.8B r15	12.4%	measured
97	Jev-Style 0.8B v3	12.0%	measured
98	Kodiak v0.2 1B (accuracy)	10.6%	measured
99	Lavoir	9.2%	measured
100	Intern-Decision-0.8B	8.4%	measured
101	jeff	7.8%	measured
102	Qwen-2.5-1B-RLCD	7.7%	measured
103	Bekko System One v0 68M	6.5%	measured
104	CLM-v0.1-8B	5.7%	measured
105	LFM2.5-2.6B-RLCD	5.4%	measured
106	Bosun v3.1 0.6B	5.1%	measured
107	DM-JEPA 1.1	2.6%	measured
108	LFM2.5-350M-RLCD	2.0%	measured
109	Zev	1.8%	measured
110	system-one-gemma	1.7%	measured
111	Julia 1	1.5%	measured
112	Bekko System One v0 17M	0.6%	measured
113	Verdict	0.0%	measured

Decision Index Language leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General