BREAKING
Grok 4.6 Lands in Devin
Third on FrontierCode Main
Fable 5
53.5
Opus 5
53.4
Grok 4.6
48
GPT-5.6 Sol
47.5
Grok 4.5
42.4
0
K
context tokens
0
reasoning levels
Longer Training, Agentic Focus
1
Supplemental training
↓
2
Regenerated SFT
↓
3
Agentic RL
↓
4
Devin availability
Promising Scores, Open Questions
Early upside
Positive
●
Strong secondary model
●
Competitive coding value
Open questions
Caution
●
Bug-for-bug validation
●
Larger context requested
Now Available in Devin
AI NEWS BLITZ
Cognition has added SpaceXAI's newest frontier model to its coding agent.