MaxxBench: Models Clustered at 74% on DeepSWE End Up 23 Points Apart After Tiny Tweaks
AaronBergman18 · x · 2026-11-08
A repost surfaces MaxxBench, a new benchmark of 89 DeepSWE tasks with small changes to instructions or starting code — same engineering problem, slightly different requirement.
Models clustered around 74% on DeepSWE end up 23 percentage points apart on MaxxBench, raising the question: good at coding, or good at DeepSWE?
More from coding & agent
- "Your agent's favorite sandbox": a meme nails how to market Exo's hidden per-minute VMs — dbreunig · 2026-11-08
- Harness leaderboard shows OpenCode lagging; author suggests Pi as faster alternative — mariofilhoml · 2026-11-08
- Developer's AI agent ships new features to his daughter's Mermaid Dots game — jacalulu · 2026-11-08
- Dev argues OpenAI-Anthropic rivalry is what keeps opencode and t3code alive — SIGKITTEN · 2026-11-08
- Reddit User: Claude Agents Deliberately Leave Work Undone to Bait Re-engagement — TimSylvester_ · 2026-11-08
- Dev Uses AI Agent to Branch, Update and Wire Two PowerShell Modules in One Pass — dfinke · 2026-11-08