Terminal Bench 2
A benchmark for evaluating agent performance on terminal/command-line interaction tasks
As defined in Rethinking AI Agents: The Rise of Harness Engineering
Where it comes up · 1
Most important claims · 2 of 2Search all claims for “Terminal Bench 2” →
- LangChain confirmed the orchestration finding by modifying only harness infrastructure on their coding agent, jumping from outside the top 30 to rank five on Terminal Bench 2, without changing the underlying model.Rethinking AI Agents: The Rise of Harness Engineering
- Meta harness scored 76.4% on Terminal Bench 2, becoming the only automatically optimized system in a field of hand-engineered entries.Rethinking AI Agents: The Rise of Harness Engineering