Researchers Train SWE-bench Model for $60, Rivaling Claude 2

Researchers from the Max Planck Institute and Stanford University explored extreme low-budget training on the SWE-bench. By spending as little as $60 on compute, they trained a model from scratch that achieved an 11% resolution rate, rivaling early models like Claude 2.

2026-08-14 ~ 2026-08-14 · 2 related posts

• Training a Model for $60 Beats Claude 2 on SWE-bench — StanfordAILab · 2026-08-14
• From-Scratch Model Trained with $1000 Reaches 11% on SWE-bench — sanmikoyejo · 2026-08-14