Terminal-Bench 4.0

Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Version 4.0 recalibrated how much time, CPU and memory each task gets, removed eight tasks and fixed nineteen, so fewer runs fail for reasons that have nothing to do with the model. Scores are not comparable with earlier versions. Higher is better.

Claude Mythos 5.1 by Anthropic holds the best Terminal-Bench 4.0 result among tracked models, at 60.9% (released Sep 1 2026). Higher scores are better on this benchmark.

1. Claude Mythos 5.1 (Anthropic) — 60.9%; 2. GPT-6 Astra (OpenAI) — 57.9%; 3. Claude Fable 5.1 (Anthropic) — 57.88%; 4. Claude Opus 5 (Anthropic) — 51.82%; 5. Claude Fable 5 (Anthropic) — 44.55%.

16 tracked models have a published Terminal-Bench 4.0 score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.

GLM-5.3 by Z.ai is the highest-ranked model with downloadable weights on Terminal-Bench 4.0, scoring 41.82% at rank 6 overall.