Data on AI Capabilities and Benchmarking

Apr. 28, 2026

GPT-5.5 Pro achieves a new high score of 159 on the Epoch Capabilities Index.

See the thread

Apr. 10, 2026

We released early results from MirrorCode, a new long-horizon SWE benchmark co-developed with METR, showing that AI can already complete some weeks-long coding tasks.

Read the report

Mar. 23, 2026

AI has solved one of the problems in FrontierMath: Open Problems, our benchmark of real research problems that mathematicians have tried and failed to solve.

Explore the post