BenchmarksSource dated 26 Oct 2026
GPT-6.1 Terra on Terminal-Bench
OpenAI reports 50.3% task success for GPT-6.1 Terra at maximum reasoning effort on Terminal-Bench 4.0. This is a vendor result with no independent reproduction or verified cross-harness ranking.
At a glance
- Model variant
- GPT-6.1 Terra
- Benchmark
- Terminal-Bench 4.0
- Score
- 50.3%
- Evaluator
- OpenAI
- Provenance
- vendor
- Sample size
- Unknown
- Observed date
- 27 Oct 2026
- Cohort basis
- Same vendor-reported table; incomplete run configuration prevents verified ranking
- Uncertainty
- {'note': 'No confidence intervals or run date supplied', 'reported': False}
Evaluation configuration
- Comparison Scope
- Single OpenAI-reported result on Terminal-Bench 4.0; comparability is unverified.
- Mode
- GPT-6.1 Terra, maximum reasoning effort; vendor launch report. Full agent harness and sampling configuration are not established in this review.