SWE-sweep Update: OpenAI, Anthropic Models Double Bug Finds at Half Cost

The SWE-sweep real-world vulnerability mining leaderboard updated: models from OpenAI and Anthropic now find twice as many bugs at half the cost, with OpenAI's Astra leading, though still missing many real vulnerabilities.

2026-10-22 ~ 2026-10-24 · 2 related posts

• OpenAI's Astra tops SWE-sweep bug-hunting leaderboard at less than half the cost of 5.6 — OfirPress · 2026-10-22
• SWE-sweep Leaderboard: Top Models Double Bug Finds at Half the Cost, Still Under 10% — stanfordnlp · 2026-10-24