International benchmarks & arenas (48 slices)
LM Arena, Epoch ECI, LiveBench, Vals Finance/Legal/Excel, Terminal-Bench, APEX, and MedHELM. If a source is partial or older than 3 months, it is explicitly flagged next to its measurement date.
Compare leading global and open-weight AI models side by side across task-specific evaluations. Every top recommendation is computed directly from verified benchmark scores — never from a static editorial list.
Models holding top ranks across multiple flagship benchmarks (Intelligence, Writing, Coding, Excel, Reasoning, Search, and Agents). A dash “—” indicates the model was not evaluated on that benchmark.
Overall leader, best value option, and strongest alternative / open-weight models.
No single “super-ranking” exists: different sources measure different capabilities (blind human preference, automated coding and reasoning tests, industry domain exams). That is why we never average Elo ratings, percentages, and Epoch ECI into a synthetic score — every tab shows its original source transparently.
LM Arena, Epoch ECI, LiveBench, Vals Finance/Legal/Excel, Terminal-Bench, APEX, and MedHELM. If a source is partial or older than 3 months, it is explicitly flagged next to its measurement date.
We separate verified consumer web-chat subscriptions from open-weight releases (self-hosted deployment) and legacy API checkpoints, and collapse duplicate reasoning-effort configurations so a single model does not flood the top of a leaderboard.
Every model displays its developer region (US/EU, China) and deployment contour: self-hosted / private cloud (open weights) or international provider cloud.