List your company

Agentic Terminal Interaction on Terminal-Bench v1

47.5Success RateClaude-Opus-4.5
3.314.77526.2537.725Jun 16, 2026

Evaluation results

MethodLinks
Claude-Opus-4.52026.0647.5
Gemini-3-Pro2026.0646.3
Kimi-K2-Instruct-09052026.0644.5
Qwen3-Coder-480B-A35B-Instruct2026.0637.5
GLM-4.72026.0636.3
GPT-5.12026.0635
LoopCoder-v2R (number of refinement loops)=22026.0634.2
LoopCoder-v2R (number of refinement loops)=32026.0630
LoopCoder-v2R (number of refinement loops)=12026.0626.3
LoopCoder-v2R (number of refinement loops)=42026.0626.3
DeepSeek-V3.22026.0623.8
Qwen3-235B-A22B-Instruct-25072026.0615
Qwen2.5-Coder-14B-Instruct2026.068.8
Seed-Coder-8B-Instruct2026.067.5
Qwen2.5-Coder-7B-Instruct2026.066.3
DeepSeek-Coder-V2-Lite-Instruct2026.065
Qwen2.5-Coder-32B-Instruct2026.065