Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
NVFP4 / INT4 tradeoffs for Llama
#1
Thought I'd post a small recipe because a lot of threads ask similar things.

Maybe useful as a starting point: Ollama on B200, keeping MoE serving in mind from the start.

Feel free to tear it apart if you see something dumb.

I care more about stability than squeezing the last possible tokens/sec out of it.

If I had to summarize the lesson so far, it would be that clean setup and observability beat clever guesses.
Reply


Messages In This Thread
NVFP4 / INT4 tradeoffs for Llama - by localpatch78 - 09-29-2026, 10:00 PM

Forum Jump:


Users browsing this thread: 1 Guest(s)