Center for AI Safety AI Overview & Benchmark Profiles

TrustTheBench

Leaderboard Agents Insights Benchmarks Models Labs Advisor Contact

Center for AI Safety

Headquarters: San Francisco, CA, USAFounded: 2022Official Website →

Creators of MMLU and Humanity's Last Exam.

Models by Center for AI Safety (0)

No models recorded for this company yet.

TrustTheBenchINDEPENDENT AI BENCHMARK OBSERVATORY

Which benchmark should you trust for what? Independent, source-verified benchmark forensics, contamination auditing, and foundation model evaluations.

Independence Guarantee: We accept no vendor sponsorships, compute subsidies, or evaluation listing fees. Every metric, contamination score, and model evaluation is strictly verified against primary literature and reproducible harnesses.

Capability Domains

Knowledge & Recall
Reasoning & Abstraction
Mathematics & Proving
Coding & Software Eng
Instruction Following
Long Context & Retrieval
Safety & Alignment
Multimodal Vision
Agentic & OS Navigation

Directory & Tools

Trust Matrix
All Benchmarks
AI Models
AI Research Labs
Cross-Model Leaderboard
Autonomous Coding Agents
Benchmark Advisor
Capability Insights
Platform Sitemap
Contact Research Team
Feedback & Audit Report

Key Frontier Suites

FrontierMath (Epoch AI)
Humanity's Last Exam (HLE)
SWE-bench Verified
LiveCodeBench (Continuous)
AIME (Invitational Math)
PutnamBench (Lean 4)
SimpleQA (OpenAI)
Terminal-Bench (Agentic CLI)

Stay Updated

New benchmark evaluations, contamination reports, and model score updates every week.

✉Get Evaluation Updates
⚑Report an Audit Issue

Privacy Policy·Terms of Use·Legal & Attribution·Sitemap