Terminal-Bench 4.0 at a glance
A maintenance-focused major release cut saturated and flawed tasks and recalibrated compute limits to reduce measurement noise.
What changed with Terminal-Bench 4.0
Terminal-Bench 4.0, posted on August 28, 2026, removed eight tasks for reasons including saturation, refusals, and publicly available solutions, and fixed 19 others. Every task now has a flat 8-hour agent timeout, with time, CPU, and memory recalibrated so frontier models rarely run out of time.
Why Terminal-Bench 4.0 matters for businesses
The maintainers say 4.0 has fewer timeouts and errors than 3.0, and it is the Terminal-Bench version used in Artificial Analysis Intelligence Index v5. For buyers, the takeaway is that agent benchmarks are maintained like software, so always check which version a published score refers to.
Sources for this story
- Terminal-Bench 4.0 (opens in a new tab)Official
Terminal-Bench · tbench.ai
- Terminal-Bench news (opens in a new tab)Official
Terminal-Bench · tbench.ai
- Announcing Artificial Analysis Intelligence Index v5 (opens in a new tab)Benchmark
Artificial Analysis · artificialanalysis.ai
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #benchmarks
- #terminal-bench
- #agents
