August 23rd.Dark Side of the MoonAnnouncing that the kimi-k2-turbo-preview model output speed has been increased to 60 Tokens per second, thanks to the tireless efforts of our engineers.Up to 100 Tokens per second.
The model is still available at a special 50% discounted price of ¥2.00 per million tokens input (cache hits), ¥8.00 per million input (cache misses), and ¥32.00 per million output, and will return to its original price on September 1st.
As 1AI previously reported, Dark Side of the Moon releases August 1 Kimi K2 high-speed version -- Kimi-K2-turbo-preview, the model parameters are the same as Kimi-K2, but the output speed is increased from 10 Tokens per second to 40 Tokens per second. kimi-K2 is a stronger code capability and better at generalized Agent tasks. Kimi K2 is a base model for MoE architecture with more code capability and better at generic agent tasks, with 1T total parameters and 32B activation parameters.