Comment on Kimi Linear: An Expressive, Efficient Attention ArchitectureparentComments−samus10moWith a Qwen3-32B-A3B (Q8) I'm getting 10-20 t/sec on KoboldAI, e.g., llama cpp. Faster than I can read, so good enough for hobby use. I expect this model to be significantly faster, but llama.cpp-based software probably doesn't support it yet.
Comments
With a Qwen3-32B-A3B (Q8) I'm getting 10-20 t/sec on KoboldAI, e.g., llama cpp. Faster than I can read, so good enough for hobby use. I expect this model to be significantly faster, but llama.cpp-based software probably doesn't support it yet.