Model2h ago
Kimi Launches 2.8 Trillion Parameter K3 Model
The 2.8 trillion parameter K3 model features a one million token context window. Internal benchmarks place its performance near GPT 5.6 Sol and Claude Fable 5, outperforming GLM 5.2. Full weights arrive by July 27. Higher pricing for K3 suggests a pivot away from the aggressive low-cost strategies common among Chinese AI firms.