Agentic · tools

VITA-Bench leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on VITA-Bench is 47.9% by Qwen3.7 Max. 71 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	K-EXAONE 2.0	55.6%	estimated ± 10.7 pp, low confidence
2	Ornith-1.0-397B	54.9%	estimated ± 10.7 pp, low confidence
3	Muse Spark 1.1	51.3%	estimated ± 9.9 pp, low confidence
4	Claude Opus 5	50.4%	estimated ± 9.9 pp, low confidence
5	Step 5 Preview	50.3%	estimated ± 9.9 pp, low confidence
6	Kimi K3	49.8%	estimated ± 9.9 pp, low confidence
7	Hy4 preview	49.6%	estimated ± 9.9 pp, low confidence
8	Gemini 3.5 Flash	49.6%	estimated ± 9.9 pp, low confidence
9	Claude Opus 4.8	49.0%	estimated ± 9.9 pp, low confidence
10	Ornith-1.5-397B	48.2%	estimated ± 9.9 pp, low confidence
11	Inkling-Small	48.0%	estimated ± 9.9 pp, low confidence
12	Qwen3.7 Max	47.9%	measured
13	Beam	47.7%	estimated ± 9.9 pp, low confidence
14	Claude Opus 4.7 (Adaptive)	47.1%	estimated ± 9.9 pp, low confidence
15	GLM-5.2	46.9%	estimated ± 9.9 pp, low confidence
16	Ornith-1.0-35B	46.9%	estimated ± 10.7 pp, low confidence
17	Kimi K2.7 Code	46.6%	estimated ± 9.9 pp, low confidence
18	Atria Dawn Preview	46.6%	estimated ± 9.9 pp, low confidence
19	GPT-5.6 Sol	46.6%	estimated ± 9.9 pp, low confidence
20	GPT-6 Astra	46.5%	estimated ± 9.9 pp, low confidence
21	Muse Glimmer 30B	46.4%	estimated ± 9.9 pp, low confidence
22	GPT-5.5 Pro	46.3%	estimated ± 9.9 pp, low confidence
23	GPT-5.5	46.3%	estimated ± 9.9 pp, low confidence
24	GPT-5.4 Pro	46.2%	estimated ± 9.9 pp, low confidence
25	Claude Mythos 5	46.1%	estimated ± 9.9 pp, low confidence
26	GPT-5.6 Terra	46.0%	estimated ± 9.9 pp, low confidence
27	MiniMax M3	45.9%	estimated ± 9.9 pp, low confidence
28	Inkling	45.8%	estimated ± 9.9 pp, low confidence
29	DeepSeek V4 Pro 0813	45.6%	estimated ± 9.9 pp, low confidence
30	Qwen3.7 Plus	45.6%	measured
31	Claude Sonnet 5	45.6%	estimated ± 9.9 pp, low confidence
32	Claude Opus 4.6	45.4%	estimated ± 9.9 pp, low confidence
33	dots3-note Preview	45.3%	estimated ± 9.9 pp, low confidence
34	GPT-5.6 Luna	45.3%	estimated ± 9.9 pp, low confidence
35	Claude Sonnet 4.6	44.9%	estimated ± 10.7 pp, low confidence
36	GLM-5.1	44.9%	estimated ± 9.9 pp, low confidence
37	GPT-5.4	44.3%	estimated ± 9.9 pp, low confidence
38	Qwen3.6 Plus	44.3%	measured
39	Ornith-1.5-35B-A3B	44.2%	estimated ± 9.9 pp, low confidence
40	Qwen3.5 397B	43.7%	measured
41	DeepSeek V4 Flash 0731	43.7%	estimated ± 9.9 pp, low confidence
42	Step 3.7 Flash	43.4%	estimated ± 9.9 pp, low confidence
43	Ling 3.0 Flash	42.1%	estimated ± 9.9 pp, low confidence
44	Qwen 3.6 Max (preview)	41.6%	estimated ± 8.2 pp, low confidence
45	MiMo-V2.5-Pro	41.0%	estimated ± 10.7 pp, low confidence
46	Muse Spark	41.0%	estimated ± 10.7 pp, low confidence
47	Ornith-1.0-9B	40.3%	estimated ± 10.7 pp, low confidence
48	Agents-A1-4B	40.3%	measured
49	Solar Pro 4	40.3%	estimated ± 9.9 pp, low confidence
50	LFM2.5-2.6B	40.1%	estimated ± 10.7 pp, low confidence
51	MiMo-V2.5	39.6%	estimated ± 10.7 pp, low confidence
52	Kimi K2.5	38.9%	estimated ± 8.2 pp, low confidence
53	GLM-5	38.8%	estimated ± 8.2 pp, low confidence
54	Agents-A1	38.8%	measured
55	Solar Open 2	38.7%	estimated ± 9.9 pp, low confidence
56	GPT-5.4 mini	38.5%	estimated ± 9.9 pp, low confidence
57	GPT-5.2	38.4%	estimated ± 9.9 pp, low confidence
58	Qwen3.6-27B	38.4%	estimated ± 8.2 pp, low confidence
59	GPT-5.4 nano	37.7%	estimated ± 9.9 pp, low confidence
60	Kimi K2.6	37.6%	estimated ± 9.9 pp, low confidence
61	Qwen3.5-122B-A10B	36.9%	estimated ± 9.9 pp, low confidence
62	Ornith-1.5-9B	36.8%	estimated ± 9.9 pp, low confidence
63	Qwen3.6-35B-A3B	35.6%	measured
64	Gemini 3.1 Pro	35.6%	estimated ± 10.7 pp, low confidence
65	MiMo-V2-Pro	35.6%	estimated ± 10.7 pp, low confidence
66	LLaDA2.2-mini	35.0%	estimated ± 10.7 pp, low confidence
67	Qwen3.5-27B	34.5%	estimated ± 9.9 pp, low confidence
68	Qwen3.5-35B-A3B	34.5%	estimated ± 9.9 pp, low confidence
69	Kimi K2.5 (Reasoning)	34.1%	estimated ± 9.9 pp, low confidence
70	GLM-5-Turbo	33.9%	estimated ± 10.7 pp, low confidence
71	LLaDA2.2-flash	32.6%	estimated ± 9.9 pp, low confidence
72	GLM-5V-Turbo	32.2%	estimated ± 10.7 pp, low confidence
73	Gemini 3 Flash	28.6%	estimated ± 10.7 pp, low confidence
74	MiniMax M2.7	28.2%	estimated ± 10.7 pp, low confidence
75	MiMo-V2-Omni	25.6%	estimated ± 10.7 pp, low confidence
76	Claude Opus 4.5	23.3%	measured
77	LongCat-Flash-Lite-Sparse	21.7%	measured
78	DeepSeek V3.2	18.5%	measured
79	Claude Sonnet 4.5	17.0%	measured
80	GLM-4.7	15.5%	measured
81	Nemotron 3 Ultra	13.1%	estimated ± 9.9 pp, low confidence
82	Nemotron 3.5 Lightning 30B A3B NVFP4	5.2%	estimated ± 9.9 pp, low confidence
83	Nemotron 3 Super 100B	2.5%	estimated ± 10.7 pp, low confidence

VITA-Bench leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General