M3 Max for local use

[byteyard]

This keeps coming up in conversation so I figured I'd start a thread.

One thing I keep noticing is that people underestimate NVFP4 until they actually try to run this stuff for more than one afternoon.

Curious whether that lines up with what other people are seeing.

[mlxgarden]

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at workstation thermals.

[labnest]

A couple things I'd check first. If you're on llama.cpp, I'd confirm the version first and then reduce variables until the behavior is obvious. Once you know the plain path works, then add the faster path back in piece by piece.

If you post your exact settings, people here can probably narrow it down pretty fast.

[rasterbench]

A couple things I'd check first. If you're on llama.cpp, I'd confirm the version first and then reduce variables until the behavior is obvious. Once y

Can you fill in a bit more context? Is this local-only or are you routing anything externally?

[mlxgarden]

[quote]A couple things I'd check first. If you're on llama.cpp, I'd confirm the version first and then reduce variables until the behavior is obvious.

Need a little more detail here. Is this local-only or are you routing anything externally?

[thetaarch43]

Did a small comparison this afternoon. On Threadripper box with vLLM, I saw basically the same behavior.

[plainember52]

Appreciate the detail. The pattern around NVFP4 has been pretty similar for me. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[quietarch]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at tokens/sec.

[cudacrate]

Thanks, that helps. The pattern around speculative decoding has been pretty similar for me. The biggest change for me was simplifying the stack instead of adding one more optimization layer.