Inference
EN
Vals AI

Terminal-Bench 1.0

State-of-the-art set of difficult terminal-based tasks

ModelsModels with a result.46
Top resultBest result on this benchmark.63.8%GPT-5.2
Top-3 spreadPoints from first to third.3.8 pts
UpdatedDate the source changed the results.12 Jan 2026

Result and cost

47 of 47 results

Results

47 results
#ModelResultThe score from the source.CostUS dollars to run one task.TimeTime to run one task.
01GPT-5.263.8% ±5.4$1.4024 min
02Claude Sonnet 4.5 Thinking61.3% ±5.4$0.6619 min
03Gemini 3 Flash Preview60.0% ±5.5$0.1262 min
04GPT-5 Codex58.8% ±5.5$1.3811 min
05Claude Opus 4.5 Thinking57.5% ±5.5$0.9615 min
05GPT-5.1-Codex57.5% ±5.5$0.12116 min
07Claude Opus 4.556.3% ±5.5$0.7443 min
08GPT-5.1-Codex-Max53.8% ±5.6$1.0212 min
09Gemini 3 Pro51.3% ±5.6$0.3188 min
10GLM-4.750.0% ±5.6$0.2239 min
10DeepSeek V3.250.0% ±5.6$0.0257 min
10Claude Haiku 4.5 Thinking50.0% ±5.6$0.266 min
13GPT-548.8% ±5.6$0.10116 min
14GPT-5.147.5% ±5.6$0.10211 min
15Claude Sonnet 4 Thinking45.0% ±5.6$2.0212 min
16Devstral 2 251243.8% ±5.5$0.4258 min
17GLM-4.642.5% ±5.5$0.2862 min
18MiniMax M2.141.3% ±5.5$0.0637 min
18Gemini 2.5 Pro41.3% ±5.5$0.6646 min
18GLM-4.541.3% ±5.5$0.1973 min
18DeepSeek V3.141.3% ±5.5$0.39111 min
22DeepSeek V3.2 (Thinking)40.0% ±5.5$0.0486 min
22Kimi K2 Thinking40.0% ±5.5$0.3352 min
22Kimi K2 Instruct40.0% ±5.5$0.3352 min
22Labs Devstral Small40.0% ±5.5$0.1355 min
26Grok 438.8% ±5.4$7.3115 min
27Kimi K2 090537.5% ±5.4$0.4332 min
28Qwen3 Maxqwen3-max36.3% ±5.4$0.4815 min
28Qwen3 Maxqwen3-max-preview36.3% ±5.4$0.7153 min
30GPT-4.133.8% ±5.3$1.295 min
31Gemini 2.5 Flash (Preview)31.3% ±5.2$0.13479 s
32GPT-5 mini30.0% ±5.1$0.36625 min
33Grok 4.1 Fast (Reasoning)28.8% ±5.1$0.00964 min
33Magistral Medium28.8% ±5.1$1.2312 min
35Grok 4 Fast (Reasoning)27.5% ±5.0$0.1882 min
36Gemini 2.5 Flash Preview Thinking26.3% ±4.9$0.1332 min
37GPT-OSS 120B22.5% ±4.7$0.03954 s
38Grok 4.1 Fast Non Reasoning21.3% ±4.6$0.0272 min
38Mistral Large 321.3% ±4.6$0.0573 min
38Gemini 2.5 Flash Thinking21.3% ±4.6$0.4484 min
41Grok Code Fast 120.0% ±4.5$0.23885 s
42Grok 4 Fast Non Reasoning18.8% ±4.4$0.31165 s
43Llama4 Maverick Instruct Basic17.5% ±4.2$0.03555 s
44Magistral Small15.0% ±4.0$0.22110 min
45DeepSeek-R113.8% ±3.9$0.362 min
46Command A6.3% ±2.7$2.505 min
46Jamba Large 1.76.3% ±2.7$4.0210 min