📊 Benchmark Tracker · latest frontier scoresOpen the full page ↗
Interactive tracker · updated 3 Nov 2026

Frontier Benchmark Tracker

Which frontier model leads on which kind of task, how fast the lead changes hands, and who measured each number: the lab that built the model, a competitor's launch table, or an independent evaluator. Every score links to its source.

⚠️ How to read this. Most numbers are reported by labs in their own launch materials; only some are independent. Scores run with different agent harnesses, effort settings or subsets are not directly comparable, so each cell shows its note and the grid prefers the same source per row. ⚠️ marks figures that reached us through a secondary source. Disclosure: Claude, made by Anthropic, helps write this site, and Claude models appear in these tables.
reported by the model's lab a competitor's launch table independent evaluator⚠️ via a secondary source★ best in row