General

Decision Index Tools 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 Tools is 83.4% 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	83.4%	measured
2	Perplexity Decider v1.2 27B	82.4%	measured
3	elephant-1	82.1%	measured
4	d1-ultra	81.8%	estimated ± 3.0 pp, medium confidence
5	Celeris-1 Decision	79.6%	estimated ± 3.0 pp, medium confidence
6	GLiDE	79.6%	estimated ± 3.0 pp, medium confidence
7	Microsoft-Decision-1	79.0%	estimated ± 3.0 pp, medium confidence
8	Torchcast Decision 27B	78.9%	measured
9	Eikos-27B	78.5%	measured
10	Mercury Decide	77.9%	estimated ± 3.0 pp, medium confidence
11	Kev 27B	77.5%	measured
12	vLLM-SR Decision 2.0 Vega 27B	77.5%	measured
13	d1	77.0%	estimated ± 3.0 pp, medium confidence
14	simple-jev · Qwen3.8-27B	75.8%	measured
15	Perplexity Decider Flash v1.2 4B	75.7%	measured
16	deck31b (Gemma 4 31B)	75.6%	measured
17	Clef	75.5%	measured
18	Quyet-1.0-Large	74.4%	measured
19	Jev 1.14	74.4%	measured
20	JEV-27B	73.9%	measured
21	Bespoke Nimble 9B v3	73.8%	measured
22	JADE	73.6%	measured
23	Drex v2.0	72.5%	measured
24	Jevfire	72.2%	measured
25	Xor 1.2	71.6%	measured
26	Drex v1.5	71.2%	measured
27	decisio 0.4.0 (Gemma 4 12B)	70.7%	measured
28	Decider chat · Gemma-4-31B	70.5%	measured
29	Decider chat · Qwen3.6-27B	69.8%	measured
30	Blink v0.3 26B-A4B (NVFP4)	69.4%	measured
31	Surogate Rune 26B-A4B v3	69.0%	measured
32	decisio 0.4.0 (Qwen3.6-35B-A3B)	66.9%	measured
33	JoshuaSP diffusiongemma (open-jev)	66.9%	measured
34	Hopper 12B (trained)	66.5%	measured
35	JPT-35B-A3B	66.4%	measured
36	GEV-26B-Decide	66.3%	measured
37	Clef-Flash	66.0%	measured
38	vLLM-SR Decision 2.0 Lux 9B	65.0%	measured
39	Lev1 3.7B	63.7%	measured
40	Jev-Omni	63.4%	measured
41	Winnow-12B Q8	62.9%	measured
42	ezjev 4B s2	62.6%	measured
43	Hopper (G) 1.2	62.2%	measured
44	Sieve 9B	62.1%	measured
45	JPT-9B	61.0%	measured
46	JevK5 9B v0.3.3	60.8%	measured
47	Clef-Omni	60.0%	measured
48	Kev 9B v2	58.7%	measured
49	vLLM-SR Decision 2.0 Nox 4B	58.4%	measured
50	lev	58.0%	measured
51	Jevstral 8B	57.6%	measured
52	JEV-9B	56.8%	measured
53	JPT-4B	56.1%	measured
54	Intern-Decision-4B	55.8%	measured
55	NeoHorse-Jev-4B	55.7%	measured
56	Intelif Qwen3-4B	54.9%	measured
57	razorback16 openjev diffusiongemma (NVFP4, vLLM)	53.9%	measured
58	Kas 4B	53.8%	measured
59	Winnow-E4B	53.5%	measured
60	Bespoke Nimble 9B v2	53.0%	measured
61	JevK5 v0.3	53.0%	measured
62	djev	52.9%	measured
63	Ryotide Qwen v0.3.1	51.7%	measured
64	Kev 4B r10	50.1%	measured
65	RSI-Jev v5.0 VL 3B	49.5%	measured
66	Gevva E4B	48.4%	measured
67	decider-35b-a3b	48.3%	measured
68	Jobe	48.2%	measured
69	Decider 4B	47.7%	measured
70	openvons	46.9%	measured
71	vLLM-SR Decision 2.0 Sol 2B	46.6%	measured
72	Solomon v1.1	45.6%	measured
73	Gevva E2B	45.0%	measured
74	EXAONE-4.0-1.2B-JEV v0.3	44.3%	measured
75	Tev1-4B-experimental	44.2%	measured
76	SemIf	39.6%	measured
77	this-that 1.2	38.4%	measured
78	metask-jev-4b	38.4%	measured
79	Intern-Decision-2B	36.9%	measured
80	open-jev (pngwn)	36.7%	measured
81	mmastrac diffusiongemma (vLLM PR 57250)	36.6%	measured
82	decider-2b	35.7%	measured
83	JPT-0.8B	34.0%	measured
84	Bosun v3.1 1.7B	31.9%	measured
85	Sieve 2B	31.7%	measured
86	Sifr 0.8B v3.1	31.3%	measured
87	Kev 0.8B r15	30.6%	measured
88	Tev1-0.8B-experimental	30.6%	measured
89	mini-jev	29.6%	measured
90	Intern-Decision-0.8B	29.2%	measured
91	vLLM-SR Decision 2.0 Eos 0.8B	29.2%	measured
92	Bosun v3.1 0.6B	26.5%	measured
93	vLLM-SR Decision 2.0 Kai 0.6B	24.4%	measured
94	Jev-Style 0.8B v3	23.8%	measured
95	Bekko System One v0 400M	18.2%	measured
96	Kodiak v0.2 1B (accuracy)	12.8%	measured
97	CLM-v0.1-8B	12.3%	measured
98	GLiNER2.5-Decide	12.3%	measured
99	Bekko System One v0 68M	11.3%	measured
100	Lavoir	9.8%	measured
101	LFM2.5-2.6B-RLCD	8.1%	measured
102	Laya	7.9%	measured
103	Qwen-2.5-1B-RLCD	7.7%	measured
104	Von 1.3	7.6%	measured
105	Bekko System One v0 17M	6.8%	measured
106	Julia 1	6.1%	measured
107	DM-JEPA 1.1	5.9%	measured
108	Verdict	5.6%	measured
109	LFM2.5-350M-RLCD	3.5%	measured
110	system-one-gemma	2.5%	measured
111	Zev	2.5%	measured
112	JevK5 Lite	0.0%	measured
113	jeff	0.0%	measured

Decision Index Tools leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General