Agentic · terminal

Terminal-Bench 2.1 (Vals AI) leaderboard

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

As of 2026-10-07, the highest measured score on Terminal-Bench 2.1 (Vals AI) is 87.3% by GPT-6 Astra. 70 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Sonnet 5.5	98.6%	estimated ± 4.6 pp, low confidence
2	Claude Opus 5.5	94.5%	estimated ± 4.6 pp, low confidence
3	Claude Mythos 5.1	89.3%	estimated ± 4.6 pp, low confidence
4	GPT-6 Astra	87.3%	measured
5	Gemini 4 Argon	86.3%	estimated ± 4.6 pp, medium confidence
6	GPT-5.6 Sol	85.8%	measured
7	Claude Fable 5.1	85.0%	measured
8	Claude Opus 5	84.6%	measured
9	Pareto 26.9	81.8%	estimated ± 4.6 pp, medium confidence
10	Pareto 26.10 Preview	81.7%	estimated ± 4.6 pp, medium confidence
11	Gemini 3.8 Flash	81.3%	measured
12	Kimi K3	80.9%	measured
13	Claude Fable 5	80.5%	measured
14	GPT-5.6 Luna	79.0%	measured
15	GPT-6.1 Sol	78.9%	estimated ± 8.5 pp, medium confidence
16	Grok 4.6	78.3%	measured
17	GPT-6 Sol	78.0%	estimated ± 8.5 pp, medium confidence
18	Ling 3.1 Flash	77.7%	estimated ± 4.6 pp, medium confidence
19	Gemini 3.7 Flash	77.5%	measured
20	GPT-5.6 Terra	77.5%	measured
21	Claude Haiku 5.5	77.4%	estimated ± 4.6 pp, medium confidence
22	MiMo-V2.6-Pro	76.8%	estimated ± 4.6 pp, medium confidence
23	Step 5 Preview	76.6%	estimated ± 4.6 pp, medium confidence
24	GPT-5.5	76.4%	measured
25	MiMo-V2.6-Flash	76.3%	estimated ± 4.6 pp, medium confidence
26	SWE-2	76.2%	estimated ± 4.6 pp, medium confidence
27	Mistral Large 4	75.6%	estimated ± 8.5 pp, medium confidence
28	Ornith-1.5-397B	75.0%	estimated ± 3.4 pp, high confidence
29	Claude Sonnet 5	74.5%	measured
30	DeepSeek V4.1 Flash	74.5%	measured
31	Gemini 3.5 Flash	74.2%	measured
32	Claude Mythos 5	73.9%	estimated ± 7.4 pp, medium confidence
33	Gemini 3.6 Flash	73.8%	measured
34	Grok 4.7	73.4%	measured
35	Muse Spark 1.3	72.3%	measured
36	Claude Opus 4.8	71.9%	measured
37	GLM-5.3	71.5%	measured
38	Gemini 3.1 Pro	70.8%	measured
39	GPT-6 Luna	70.3%	estimated ± 8.5 pp, medium confidence
40	Muse Spark 1.2	69.7%	measured
41	Sakana Fugu-Ultra	69.4%	estimated ± 7.4 pp, medium confidence
42	Muse Spark 1.1	69.3%	measured
43	Ember-1	69.3%	estimated ± 7.4 pp, medium confidence
44	SWE-1.7	68.9%	estimated ± 7.4 pp, medium confidence
45	Claude Opus 4.7	68.5%	measured
46	GPT-5.3 Codex	68.0%	estimated ± 6.3 pp, medium confidence
47	Sakana Fugu	67.9%	estimated ± 7.4 pp, medium confidence
48	Beam	67.8%	estimated ± 7.4 pp, medium confidence
49	GLM-5.2	67.8%	measured
50	Grok 4.5	67.8%	measured
51	Qwen3.8 Max	67.4%	measured
52	DeepSeek V4 Flash 0731	67.0%	measured
53	Kimi K2.7 Code	67.0%	measured
54	Atria Dawn Preview	66.4%	estimated ± 7.4 pp, medium confidence
55	GPT-5.4	66.0%	estimated ± 6.3 pp, medium confidence
56	Ornith-1.0-397B	65.8%	estimated ± 7.4 pp, medium confidence
57	dots3-note Preview	64.0%	estimated ± 7.4 pp, medium confidence
58	GLM-5.3-Flash	62.9%	measured
59	Ornith-1.5-35B-A3B	61.7%	estimated ± 3.4 pp, high confidence
60	Qwen3.7 Max	61.0%	measured
61	Claude Opus 4.7 (Adaptive)	61.0%	estimated ± 6.3 pp, medium confidence
62	Seed 2.1 Pro	60.8%	estimated ± 7.4 pp, medium confidence
63	MiMo-V2.5	60.7%	measured
64	Apodex 1.1	60.6%	estimated ± 7.4 pp, medium confidence
65	Laguna S 2.1	60.2%	estimated ± 7.4 pp, medium confidence
66	Quasar 438B	59.5%	estimated ± 7.4 pp, medium confidence
67	Composer 2.5	58.4%	measured
68	Qwen3.8-27B	58.4%	measured
69	Seed 2.1 Turbo	58.2%	estimated ± 7.4 pp, medium confidence
70	Claude Opus 4.6	57.4%	estimated ± 6.3 pp, medium confidence
71	Qwen 3.6 Max (preview)	57.4%	estimated ± 6.3 pp, medium confidence
72	Claude Sonnet 4.6	57.3%	measured
73	MiMo-V2.5-Pro	57.3%	measured
74	GLM-5.1	56.9%	measured
75	Pokee-Isaac 28B	56.2%	estimated ± 7.4 pp, medium confidence
76	Ornith-1.0-35B	55.6%	estimated ± 7.4 pp, medium confidence
77	Hy4 preview	55.1%	measured
78	Inkling-Small	55.1%	measured
79	DeepSeek V4 Pro 0813	54.7%	measured
80	GPT-5.4 mini	54.7%	measured
81	MAI-Code-1.1-Flash	54.5%	estimated ± 7.4 pp, medium confidence
82	Composer 2	54.1%	estimated ± 6.3 pp, medium confidence
83	Gemini 3 Flash	53.9%	measured
84	Kimi K2.6	53.6%	measured
85	MiniMax M3	53.6%	measured
86	Qwen3.6 Plus	53.2%	measured
87	Qwen3.7 Plus	52.8%	measured
88	Claude Opus 4.5	52.0%	estimated ± 6.3 pp, medium confidence
89	Qwen3.6-27B	52.0%	estimated ± 6.3 pp, medium confidence
90	Step 3.7 Flash	51.9%	estimated ± 7.4 pp, medium confidence
91	Muse Spark	51.7%	estimated ± 6.3 pp, medium confidence
92	Muse Glimmer 30B	51.1%	estimated ± 6.0 pp, low confidence
93	Nemotron 3 Ultra	50.9%	measured
94	Gemini 3.5 Flash-Lite	50.2%	measured
95	Ling 3.0 Flash	50.2%	measured
96	Solar Pro 4	50.0%	estimated ± 7.4 pp, medium confidence
97	GLM-5	49.2%	estimated ± 6.3 pp, medium confidence
98	MiniMax M2.7	48.7%	measured
99	Hy3 Preview	47.6%	estimated ± 6.3 pp, medium confidence
100	Inkling	47.6%	measured
101	Ternary Bonsai 2 27B	46.7%	estimated ± 7.4 pp, low confidence
102	Qwen3.5 397B	45.9%	estimated ± 6.3 pp, medium confidence
103	Qwen3.6-35B-A3B	45.0%	estimated ± 6.3 pp, medium confidence
104	Kimi K2.5	44.4%	estimated ± 6.3 pp, medium confidence
105	Kimi K2.5 (Reasoning)	44.4%	estimated ± 6.3 pp, medium confidence
106	Grok 4.20	44.2%	measured
107	Claude Haiku 4.5	43.8%	measured
108	Claude Sonnet 4.5	43.7%	estimated ± 6.3 pp, medium confidence
109	Qwen3.5-122B-A10B	43.2%	estimated ± 6.3 pp, medium confidence
110	Grok 4.3	41.9%	measured
111	Ornith-1.5-9B	41.6%	estimated ± 7.4 pp, low confidence
112	GPT-5.4 nano	41.6%	measured
113	MAI-Thinking-1	40.2%	estimated ± 6.3 pp, medium confidence
114	K-EXAONE 2.0	39.8%	estimated ± 7.4 pp, low confidence
115	Ornith-1.0-9B	39.3%	estimated ± 7.4 pp, low confidence
116	Mistral Medium 3.5 128B	39.0%	measured
117	Qwen3.5-27B	36.2%	estimated ± 6.3 pp, medium confidence
118	GLM-4.7	35.7%	estimated ± 6.3 pp, medium confidence
119	Qwen3.5-35B-A3B	35.3%	estimated ± 6.3 pp, medium confidence
120	Mercury 2.5	34.5%	measured
121	Gemini 3.1 Flash-Lite	34.1%	measured
122	Laguna M.1	34.1%	measured
123	A.X K2	33.8%	estimated ± 7.4 pp, low confidence
124	Laguna XS 2.1	32.6%	estimated ± 6.3 pp, medium confidence
125	LongCat-Flash-Lite-Sparse	29.2%	estimated ± 6.3 pp, low confidence
126	Granite 4.2 30B	28.5%	estimated ± 7.4 pp, low confidence
127	Laguna XS.2	25.8%	measured
128	Nemotron 3.5 Lightning 30B A3B NVFP4	24.1%	estimated ± 7.4 pp, low confidence
129	Granite 4.2 8B	21.8%	estimated ± 7.4 pp, low confidence
130	Command A+	17.6%	measured
131	MiniCPM5-2B	12.6%	estimated ± 7.4 pp, low confidence

Terminal-Bench 2.1 (Vals AI) leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General