Grok Build 4.7 takes the lead on DeepSWE benchmark, topping OpenAI's Astra

Kuprel · x · 2026-09-24

Per Artificial Analysis's DeepSWE benchmark, Grok Build 4.7 has taken the lead, scoring above OpenAI's Astra at modifying repos. The author questions whether xAI's coding agent now genuinely leads and tags @elonmusk for confirmation.

Original post →

More from Models

Models channel →