General

Decision Index Full Score 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 Full Score is 68.4% 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	68.4%	measured
2	Perplexity Decider v1.2 27B	68.0%	measured
3	d1-ultra	63.6%	estimated ± 4.5 pp, medium confidence
4	deck31b (Gemma 4 31B)	63.0%	measured
5	Decider chat · Gemma-4-31B	61.6%	measured
6	Torchcast Decision 27B	61.5%	measured
7	Quyet-1.0-Large	60.8%	measured
8	elephant-1	60.3%	measured
9	Jev 1.14	60.2%	measured
10	Kev 27B	59.6%	measured
11	Celeris-1 Decision	59.3%	estimated ± 4.5 pp, medium confidence
12	GLiDE	59.3%	estimated ± 4.5 pp, medium confidence
13	Eikos-27B	58.8%	measured
14	Microsoft-Decision-1	58.4%	estimated ± 4.5 pp, medium confidence
15	vLLM-SR Decision 2.0 Vega 27B	58.0%	measured
16	JEV-27B	57.4%	measured
17	Mercury Decide	57.1%	estimated ± 4.5 pp, medium confidence
18	simple-jev · Qwen3.8-27B	57.0%	measured
19	Surogate Rune 26B-A4B v3	56.4%	measured
20	d1	56.1%	estimated ± 4.5 pp, medium confidence
21	Blink v0.3 26B-A4B (NVFP4)	55.9%	measured
22	JADE	55.6%	measured
23	Clef	54.9%	measured
24	Decider chat · Qwen3.6-27B	54.7%	measured
25	Perplexity Decider Flash v1.2 4B	54.3%	measured
26	JPT-35B-A3B	54.1%	measured
27	Xor 1.2	53.6%	measured
28	GEV-26B-Decide	53.5%	measured
29	Drex v2.0	52.8%	measured
30	Bespoke Nimble 9B v3	51.1%	measured
31	decisio 0.4.0 (Qwen3.6-35B-A3B)	50.8%	measured
32	JoshuaSP diffusiongemma (open-jev)	49.5%	measured
33	decisio 0.4.0 (Gemma 4 12B)	48.8%	measured
34	decider-35b-a3b	46.6%	measured
35	Winnow-12B Q8	46.6%	measured
36	vLLM-SR Decision 2.0 Lux 9B	46.0%	measured
37	Hopper 12B (trained)	45.8%	measured
38	Drex v1.5	45.8%	measured
39	Clef-Omni	45.0%	measured
40	JevK5 9B v0.3.3	44.9%	measured
41	Clef-Flash	44.4%	measured
42	Jevfire	44.4%	measured
43	JPT-9B	44.1%	measured
44	ezjev 4B s2	43.9%	measured
45	Jev-Omni	43.8%	measured
46	Solomon v1.1	43.2%	measured
47	Bespoke Nimble 9B v2	42.1%	measured
48	JEV-9B	42.1%	measured
49	Sieve 9B	41.7%	measured
50	Kev 9B v2	40.9%	measured
51	vLLM-SR Decision 2.0 Nox 4B	40.8%	measured
52	Hopper (G) 1.2	40.2%	measured
53	JPT-4B	39.6%	measured
54	Intern-Decision-4B	38.7%	measured
55	RSI-Jev v5.0 VL 3B	38.6%	measured
56	djev	38.1%	measured
57	razorback16 openjev diffusiongemma (NVFP4, vLLM)	37.6%	measured
58	NeoHorse-Jev-4B	37.3%	measured
59	JevK5 v0.3	36.6%	measured
60	Decider 4B	36.3%	measured
61	lev	35.5%	measured
62	Lev1 3.7B	35.4%	measured
63	Kev 4B r10	34.9%	measured
64	mmastrac diffusiongemma (vLLM PR 57250)	34.3%	measured
65	Kas 4B	34.0%	measured
66	Tev1-4B-experimental	33.7%	measured
67	Winnow-E4B	32.7%	measured
68	Jevstral 8B	32.2%	measured
69	metask-jev-4b	32.2%	measured
70	Ryotide Qwen v0.3.1	32.0%	measured
71	Intelif Qwen3-4B	31.3%	measured
72	open-jev (pngwn)	31.1%	measured
73	Jobe	30.4%	measured
74	Gevva E4B	29.2%	measured
75	openvons	28.8%	measured
76	SemIf	26.7%	measured
77	this-that 1.2	24.9%	measured
78	Gevva E2B	24.3%	measured
79	decider-2b	23.6%	measured
80	vLLM-SR Decision 2.0 Sol 2B	23.6%	measured
81	mini-jev	22.8%	measured
82	Sieve 2B	21.9%	measured
83	Intern-Decision-2B	20.9%	measured
84	EXAONE-4.0-1.2B-JEV v0.3	20.6%	measured
85	vLLM-SR Decision 2.0 Eos 0.8B	18.4%	measured
86	Bekko System One v0 400M	18.0%	measured
87	Bosun v3.1 1.7B	16.5%	measured
88	vLLM-SR Decision 2.0 Kai 0.6B	16.4%	measured
89	Kev 0.8B r15	16.3%	measured
90	JPT-0.8B	15.5%	measured
91	Sifr 0.8B v3.1	14.5%	measured
92	Kodiak v0.2 1B (accuracy)	13.9%	measured
93	Intern-Decision-0.8B	12.7%	measured
94	Jev-Style 0.8B v3	12.5%	measured
95	Tev1-0.8B-experimental	12.5%	measured
96	Bekko System One v0 68M	12.4%	measured
97	Von 1.3	11.7%	measured
98	Bosun v3.1 0.6B	11.0%	measured
99	Lavoir	10.8%	measured
100	GLiNER2.5-Decide	10.1%	measured
101	Laya	9.2%	measured
102	JevK5 Lite	8.8%	measured
103	LFM2.5-2.6B-RLCD	8.2%	measured
104	Qwen-2.5-1B-RLCD	7.8%	measured
105	Bekko System One v0 17M	7.7%	measured
106	CLM-v0.1-8B	6.5%	measured
107	jeff	6.2%	measured
108	LFM2.5-350M-RLCD	5.9%	measured
109	system-one-gemma	5.2%	measured
110	DM-JEPA 1.1	5.0%	measured
111	Julia 1	4.8%	measured
112	Zev	3.4%	measured
113	Verdict	2.0%	measured

Decision Index Full Score leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General