Inference
EN
Vals AI

Terminal-Bench 2.0

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

ModelsModels with a result.66
Top resultBest result on this benchmark.73.2%GPT-5.5
Top-3 spreadPoints from first to third.4.7 pts
UpdatedDate the source changed the results.4 Jun 2026

Result and cost

63 of 66 results

Results

66 results
#ModelResultThe score from the source.CostUS dollars to run one task.TimeTime to run one task.
01GPT-5.573.2% ±4.0$0.7618 min
02Claude Opus 4.870.0% ±1.4$0.8848 min
03Claude Opus 4.768.5% ±5.0$0.97611 min
04Gemini 3.5 Flash67.4% ±1.3$1.197 min
04Gemini 3.1 Pro Preview67.4% ±5.0$0.5169 min
06GPT-5.3 Codex64.0% ±5.1$0.3458 min
07Muse Spark59.6% ±5.2$0.079 min
07Claude Sonnet 4.659.6% ±5.2$0.51910 min
09Qwen3.7 Max59.2% ±1.4$0.71911 min
10GPT-5.458.4% ±5.3$0.51217 min
10Claude Opus 4.6 (Adaptive)58.4% ±5.3$1.1415 min
10Claude Opus 4.558.4% ±5.3$1.1810 min
13Kimi K2.657.3% ±5.3$0.19313 min
14DeepSeek V4 Pro 042356.2% ±5.3$0.1814 min
15Gemini 3 Pro55.1% ±5.3$0.4567 min
16Claude Opus 4.5 Thinking53.9% ±5.3$1.1211 min
16GLM-5.153.9% ±5.3$0.15815 min
18GPT-5.251.7% ±5.3$0.4712 min
18Gemini 3 Flash Preview51.7% ±5.3$0.1698 min
18Qwen 3.6 Max (preview)51.7% ±5.3$0.3215 min
21GLM-5 Thinking49.4% ±5.3$0.24715 min
22MiniMax M2.747.2% ±5.3$0.07414 min
23MiniMax M346.1% ±5.3$0.21617 min
24GPT-5.144.9% ±5.3$0.40815 min
24Qwen3.6 Plus44.9% ±5.3$0.14614 min
24GPT-5.4 mini44.9% ±4.5$0.30512 min
24Qwen3.6-27B44.9% ±5.3$0.29833 min
28Grok 4.343.4% ±4.4$0.37311 min
29Claude Sonnet 4.5 Thinking41.6% ±5.3$0.75112 min
29MiniMax M2.541.6% ±5.3$0.08513 min
29Qwen3.5 Plus Thinking41.6% ±5.3$0.32231 min
32Grok 4.20 (Reasoning)40.4% ±5.2$0.1543 min
32Kimi K2.5 Thinking40.4% ±5.2$0.1413 min
34GPT-5.4 nano39.9% ±4.5$0.07514 min
35Gemma 4 31B39.3% ±5.2—14 min
36GLM-4.738.2% ±5.2$0.18413 min
36Claude Haiku 4.5 Thinking38.2% ±5.2$0.32811 min
38GPT-537.1% ±5.1$0.47716 min
38MiniMax M2.137.1% ±5.1$0.0713 min
38Kimi K2 Thinking37.1% ±5.1$0.11616 min
41DeepSeek V3.2 (Thinking)36.0% ±5.1$0.16617 min
42DeepSeek V3.234.8% ±5.1$0.06313 min
43Laguna M.131.5% ±5.0—0.0 s
44Gemini 2.5 Pro30.3% ±4.9$0.47213 min
44Mistral Medium 3.530.3% ±4.9$4.7121 min
46Grok 4 Fast (Reasoning)29.2% ±4.8$0.0347 min
47Grok 428.1% ±4.8$0.55315 min
47GLM-4.628.1% ±4.8$0.14410 min
47Laguna XS.228.1% ±4.8—0.0 s
50GPT-5 mini27.0% ±4.7$0.11717 min
51Kimi K2 Instruct25.8% ±4.7$1.1112 min
52Gemini 3.1 Flash Lite Preview24.7% ±4.6$0.06511 min
52Grok 4.1 Fast (Reasoning)24.7% ±4.6$0.0568 min
52Qwen3 Max24.7% ±4.6$0.42315 min
52Qwen3.5 Flash24.7% ±4.6$0.1112 min
56DeepSeek V3.122.5% ±4.4$0.50214 min
57Gemini 2.5 Flash Preview Thinking21.3% ±4.4$0.1447 min
58GPT-OSS 120B19.1% ±4.2$0.1328 min
59Trinity-Large-Thinking18.0% ±4.1$0.32617 min
59Grok 4.1 Fast Non Reasoning18.0% ±4.1$0.0334 min
61Mistral Small 416.9% ±4.0$0.1067 min
62GPT-4.114.6% ±3.8$1.0810 min
63Magistral Medium13.5% ±3.6$0.56715 min
64Mistral Large 39.0% ±3.0$0.0320 min
65Llama4 Maverick Instruct Basic2.2% ±1.6$0.03420 min
65Command A2.2% ±1.6$2.4413 min