Model parameters confusion
#67
by Ujjwal-Tyagi - opened
Kimi K2.5 is 171B paramters but in model card, there is 1T parameters mentioned, it might be confusing for others, Model Summary in the model card is kinda wrong, please check it out!
Ujjwal-Tyagi changed discussion status to closed
Ujjwal-Tyagi changed discussion status to open
Kimi K2.5 has 1 trillion parameters with 32 billion active. HF shows 171B probably because the model was trained using INT4 just like Kimi-K2-Thinking. If you search for Kimi K2 Instruct, it shows 1T, they didn't change the base model for k2.5.