Claude Opus 5 hits 74% on DeepSWE, topping long-horizon coding models

brandon_galang · x · 2026-07-29

The poster says Opus 5 is “objectively cracked” and should be studied by everyone using it, citing a quote that Claude Opus 5 has reached 74% on DeepSWE and is the best long-horizon coding model seen so far. The core claim is that Anthropic’s latest model is now leading this benchmark for long-horizon coding tasks.

Original post →

More from Models

Models channel →