New
AI & RoboticsMore in AI & Robotics→
Kimi K3 lags top U.S. models in cyber exploit tests
Key Takeaways
- A joint U.K.-U.S. evaluation found Kimi K3 could assist offensive cyber tasks without meaningful resistance.
- Kimi K3 scored 32.2 percent on ExploitBench versus 76.2 percent for leading U.S. models.
- The model outperformed GLM-5.2 but failed to reach Arbitrary Code Execution on any task.
- The findings suggest Chinese models are improving on cyber tasks while still trailing the U.S. frontier.
DE
DT Editorial Team··via the-decoder.com