List your company

Terminal-based problem solving on Terminal-Bench 2 (out-of-distribution)

34.12Task Success RateCODESKILL
19.435223.247627.0630.8724May 25, 2026

Evaluation results

MethodLinks
CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=Qwen3.5-35B-A3B2026.0534.1230.29
Subtask MemorySkill Backbone=GPT-5.4-mini, Frozen coding policy=Qwen3.5-35B-A3B2026.0530.5937.88
Prompt Skill Mgmt.Skill Backbone=GPT-5.4-mini, Frozen coding policy=Qwen3.5-35B-A3B2026.0528.2433.38
No-skill baselineSkill Backbone=-, Frozen coding policy=Qwen3.5-35B-A3B2026.0525.8844.05
CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=GPT-5.4-mini2026.0525.887.8
Prompt Skill Mgmt.Skill Backbone=Qwen3.5-4B, Frozen coding policy=Qwen3.5-35B-A3B2026.0524.7139.73
SFT-CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=Qwen3.5-35B-A3B2026.0524.7134.62
Prompt Skill Mgmt.Skill Backbone=Qwen3.5-4B, Frozen coding policy=GPT-5.4-mini2026.0523.536.25
SFT-CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=GPT-5.4-mini2026.0523.536.5
Subtask MemorySkill Backbone=GPT-5.4-mini, Frozen coding policy=GPT-5.4-mini2026.0522.356.06
Prompt Skill Mgmt.Skill Backbone=GPT-5.4-mini, Frozen coding policy=GPT-5.4-mini2026.0521.186.06
No-skill baselineSkill Backbone=-, Frozen coding policy=GPT-5.4-mini2026.05208.23