Terminal-Bench 2.0 on claude-opus-4-5
89 tasks, 1 attempt, open network. Updated 25 Oct 2026.
Compare
Pick two harnesses to compare them task by task.
OpenCode against mux, task by task
Over the 89 tasks both ran, 1 attempt each
OpenCode minus mux, pass rate
−6.7 points
95% interval −15.7 to +2.2
The interval includes zero, so the paired test can't tell the two apart.
Where the tasks landed
- 6Only OpenCode solved
- 40Both solved
- 12Only mux solved
- 31Neither solved
Of 89 tasks, 6 only OpenCode solved, 12 only mux solved, 40 both solved and 31 neither did.
| Task | Reward | Tokens | Time | |||
|---|---|---|---|---|---|---|
| OpenCode | mux | OpenCode | mux | OpenCode | mux | |
break-filter-js-from-html | 0 | 1 | — | — | 20.0 min | 8.2 min |
build-pmars | 0 | 1 | — | 4,799 | 2.1 min | 1.9 min |
cancel-async-tasks | 0 | 1 | — | 2,307 | 22 s | 45 s |
configure-git-webserver | 0 | 1 | — | — | 2.6 min | 2.0 min |
count-dataset-tokens | 0 | 1 | — | — | 1.9 min | 2.0 min |
db-wal-recovery | 0 | 1 | — | 3,046 | 6.4 min | 3.4 min |
dna-assembly | 1 | 0 | — | — | 14.7 min | 15.6 min |
extract-moves-from-video | 1 | 0 | — | — | 21.3 min | 17.1 min |
financial-document-processor | 0 | 1 | — | — | 1.2 min | 2.8 min |
largest-eigenval | 1 | 0 | — | — | 6.5 min | 3.5 min |
overfull-hbox | 1 | 0 | — | — | 5.0 min | 3.7 min |
path-tracing-reverse | 0 | 1 | — | — | 19.1 min | 13.6 min |
pypi-server | 1 | 0 | — | 7,463 | 2.7 min | 3.4 min |
pytorch-model-recovery | 0 | 1 | — | — | 3 s | 9.7 min |
qemu-startup | 0 | 1 | — | 3,945 | 15.0 min | 3.5 min |
rstan-to-pystan | 1 | 0 | — | — | 12.1 min | 30.0 min |
sanitize-git-repo | 0 | 1 | — | — | 1.9 min | 3.2 min |
sparql-university | 0 | 1 | — | — | 37 s | 1.6 min |