the terminal-bench 4.1 landscape
more tasks passed.
less spend.
updated · 12 nov 2026
published frontier models and effort levels. find the point where capability earns its cost.
terminal-bench 4.1 · total benchmark cost in usd
77published variants
73plottable results
14on this frontier
4.1terminal-bench
the efficient edge
higher is smarter. further left is cheaper.
how to read the edge
nothing better.
for less.
a point is on the frontier when no other selected variant is at least as accurate and at least as cheap, with one strict improvement. equal results share a place.
the complete dataset 83 variants · source links included
| model / effort | score (%) ↑ | total cost ↓ | selected frontier | source |
|---|