BREAKING
Cognition Adds Claude Sonnet 5 to Devin
Sonnet 5 vs Opus 4.8
Sonnet 5New
Frontier-level coding
Lower cost tier
Backend for Devin
Opus 4.8High-end
Top-tier model
Extended score 51.8%
Beaten on FrontierCode
FrontierCode Extended Tasks
Extended150
Main100
Diamond50
How Devin Works Autonomously
1Plan
2Execute
3Test
4Debug
5Pull request
0%
Planning gain
0%
End-to-end eval
0x
Speedup
Real-World Verdict Still Pending
AI NEWS BLITZ
Cognition just brought Anthropic's Claude Sonnet 5 to its AI software engineer Devin.