SWE-1.7
SWE-1.7, post-trained from Kimi K2.7, scores 42.3% on FrontierCode 1.1 Main and runs at 1,000 tok/s on Cerebras.
Cognition keeps building its models on open-weights Chinese bases plus RL. SWE-1.6 had scored 9.4% on the same benchmark, so the jump is large; Terminal-Bench 2.1 81.5% and SWE-Bench Multilingual 77.8%.
- Date
- Wednesday, 8 July 2026
- Lab
- Cognition
- Kind
- model
- Access
- app only
Figures
| Measure | Value | Measured by |
|---|---|---|
| FrontierCode 1.1 Main | 42.3% SWE-1.6 9.4% | company |
| Terminal-Bench 2.1 | 81.5% | company |
All scores are Cognition's own; the Kimi K2.7 base is stated in the post.
Sources
This record was checked against its sources on 6 October 2026. How we check