BREAKING
Cognition Unveils SWE-1.7 Coding Model
Built on Kimi K2.7 Plus Custom RL
1
Kimi K2.7 base
↓
2
Stabilized RL
↓
3
Self-compaction
↓
4
SWE-1.7
FrontierCode 1.1 Main Scores
SWE-1.7
42.3
Kimi K2.7
30.1
GPT-5.5
43
Opus 4.8
46.5
0
tok/s
via Cerebras
0
$
per task
Praise and Open Questions
Upside
●
1,000 tok/s speed
●
Near-GPT-5.5 at lower cost
Concerns
●
Possible benchmaxing
●
Awaiting real-world tests
Available Now Inside Devin
AI NEWS BLITZ
Cognition has launched SWE-1.7, a new model for autonomous coding agents.