4 ms·

How many tokens per second?

• LuxBennu 7mo ago
  Roughly 8-12 token/s on generation depending on context length. Prompt processing is faster obviously. Haven't benchmarked it super carefully though, just eyeballing the llama.cpp output.