AnalysisAI ModelsJuly 24, 2026
Kimi K3 vs Claude Fable 5 on DeepSWE: Cost and Coding
Together AI ran 452 DeepSWE rollouts on the two models. Claude Fable 5 leads pass@1 by 1.4 points, but Kimi K3 wins pass@4 and delivers 2.8x the solves per dollar.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Meta working on voice dictation for Meta AI website
- Gallup: The more Americans know about AI, the less they like it
- AgentSys orchestrates 49 agents across five IDEs
- 5.6 sol rolled out on Cerebras for select customers
- Codex user: Work quality 10x'd in 3 months with AI voice mode