The UK AI Security Institute is publishing verified frontier-model results through EvalEval's shared schema and Evaluation Cards. The first release covers six frontier models across HealthBench, FrontierMath, Humanity's Last Exam, SWE-Bench Pro, Terminal-Bench 2.0 and two cyber evaluations. Each record carries configuration context and evaluation methods instead of only a leaderboard number. The publication also examines how inference-time compute and protocol choices change reported performance.
UK AISI publishes reproducible frontier evaluations through EvalEval
2026-09-22Verified announcementWATCH
The UK AI Security Institute is publishing verified frontier-model results through EvalEval's shared schema and Evaluation Cards. The first release covers six frontier models across HealthBench, FrontierMath, Humanity's Last Exam, SWE-Bench Pro, Terminal-Bench 2.0 and two cyber evaluations. Each record carries configuration context and evaluation methods instead of only a leaderboard number. The publication also examines how inference-time compute and protocol choices change reported performance.
Imported from the existing research archive. Migration preserves the recorded judgment and is not a fresh verification.
What this means for you
For teams comparing model claims, a machine-readable result with configuration and method is much easier to audit than a screenshot or vendor table. This is concrete government-evaluator adoption of a shared reporting format, but it is still one publisher. Watch for other labs and evaluators using the same schema before treating it as a standard.
Evidence & related material
Technical details & scores
| Format | EvalEval shared schema plus Evaluation Cards |
|---|---|
| Publisher | UK AI Security Institute |
| Coverage | six frontier models; seven evaluation suites |
| Benchmarks | HealthBench, FrontierMath, Humanity's Last Exam, SWE-Bench Pro, Terminal-Bench 2.0 and two cyber evaluations |
| Reported Context | model configuration, inference-time compute and evaluation protocol |
Technical impact 3
Engineering value 3
Adoption signal 2
Maturity 3
Verification cost 2
Risk 1