Terminal-Bench Hard

Artificial Analysis · 50 models

Rank	Model	Score
1	GPT-5.6 Sol (Max)	0.6591
2	Claude Fable 5 (Max, Opus 4.8 Fallback)	0.6288
2	GPT-5.6 Sol (Medium)	0.6288
2	GPT-5.6 Terra (Xhigh)	0.6288
5	GPT-5.6 Sol (High)	0.6212
6	GPT-5.6 Sol (Xhigh)	0.6136
7	GPT-5.5 (Xhigh)	0.6061
7	GPT-5.6 Sol (Low)	0.6061
9	GPT-5.5 (High)	0.5985
10	Claude Opus 4.8 (Max)	0.5833
11	GPT-5.4 (Xhigh)	0.5758
11	GPT-5.5 (Medium)	0.5758
11	GPT-5.6 Terra (High)	0.5758
11	GPT-5.6 Terra (Max)	0.5758
15	Claude Opus 4.7 (Non-reasoning, High)	0.5455
16	Gemini 3.1 Pro Preview	0.5379
17	Claude Sonnet 4.6 (Max)	0.5303
17	GPT-5.3 Codex (Xhigh)	0.5303
19	GPT-5.4 mini (Xhigh)	0.5227
19	GPT-5.5 (Low)	0.5227
21	Claude Opus 4.7 (Max)	0.5152
22	GLM-5.2 (Max)	0.5076
22	Qwen3.7 Max	0.5076
24	GPT-5.5 (Non-reasoning)	0.4924
24	KAT Coder Pro V2	0.4924
26	Claude Opus 4.6 (Non-reasoning, High)	0.4848
27	Claude Opus 4.5 (Reasoning)	0.4697
27	GPT-5.2 (Xhigh)	0.4697
27	Qwen3.7 Plus	0.4697
30	Claude Opus 4.6 (Max)	0.4621
30	Claude Sonnet 4.6 (Non-reasoning, High)	0.4621
30	DeepSeek V4 Pro 0424 (Max)	0.4621
30	Gemini 3.5 Flash (Minimal)	0.4621
34	GPT-5.1 (High)	0.4545
34	Muse Spark	0.4545
36	Kimi K2.7 Code	0.447
37	GPT-5.6 Terra (Low)	0.4394
37	Kimi K2.6 (Reasoning)	0.4394
37	Qwen3.6 Max Preview	0.4394
37	Qwen3.6 Plus	0.4394
41	GLM-5 (Reasoning)	0.4318
41	GLM-5.1 (Reasoning)	0.4318
41	GPT-5.2 (Medium)	0.4318
41	GPT-5.4 (Low)	0.4318
41	MiMo-V2.5-Pro (Reasoning)	0.4318
46	Claude Sonnet 4.6 (Non-reasoning, Low)	0.4242
46	GPT-5.4 nano (Xhigh)	0.4242
46	GPT-5.5 Instant (May 2026)	0.4242
46	MiniMax-M3	0.4242
50	DeepSeek V4 Pro 0424 (High)	0.4167