Coding
Terminal-Bench 4.0
End-to-end tasks in a real terminal — install, build, debug, run — scored on whether the machine ends up in the required state.
Terminal-Bench 4.0 is scored as a percentage of questions answered correctly. 3 of the 159 models in this index report a score for it. The highest published figure here is 60.9%, from Claude Mythos 5.1.
Reported scores
Top 3 on Terminal-Bench 4.0
Ordered by the figure each maker published. Models that have not reported this benchmark are not listed — an absent score is not a low score.
| # | Model | Published score |
|---|---|---|
| 1 | Claude Mythos 5.1Anthropic | |
| 2 | Claude Fable 5.1Anthropic | |
| 3 | Claude Fable 5Anthropic |
Where these numbers come from
Every score on this page is a published figure, taken from the model's own card, system card, technical report or release post, or from a public leaderboard. CorX Labs did not run these evaluations. Most are self-reported by the lab that built the model, which means they were produced under that lab's own choice of prompt, scaffold and number of attempts — so treat them as a starting point for a shortlist, not as a settled ranking.
A score someone other than the model's maker measured is marked Independent and names its measurer. Those are the stronger numbers on this page — an outside harness has no reason to flatter anyone — and there are not many of them.
Where a figure has not been published, the cell reads Not reported rather than an estimate. Nothing here is inferred, interpolated or guessed. Each model records the month its row was last checked. Full method and caveats.
How to read this score
Terminal-Bench scores whether the machine ended up in the required state after a real session — install, build, debug, run. It is the closest thing here to watching a model actually do the job, rather than produce a patch someone else applies.
Where it is weak
The environment is doing a lot of the work, so results move with the scaffold, the timeout and the tool set. Two labs reporting Terminal-Bench may not be running comparable harnesses at all.