What worked for me was pretty simple. If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious. Not glamorous advice, but repeatable baselines save a lot of time.

Thanks, that helps. I don't think your conclusion is far off. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

There are a few different conversations hiding inside this thread, so being explicit about the goal helps.

I'd approach it in a boring order. On 4090, the easiest win is often to back off the aggressive settings and see whether the problem survives.

I tested this tonight. I couldn't reproduce the slowdown until I pushed context much higher. The main thing that moved was latency, not quality.

What exact setup are you running? Are you testing this on Threadripper box or something else?

Thanks, that helps. I ended up in about the same place using TensorRT-LLM. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

That matches what I was seeing. I don't think your conclusion is far off. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at consumer GPUs. Not glamorous advice, but repeatable baselines save a lot of time.

Did a bit of testing and thought I'd post the result here.

I tried this with Phi using SGLang on 4090. The headline result was better than I expected, but workstation thermals still ended up being the limiting factor.

Curious if anyone else landed in the same place.

Maybe, but I had the opposite result. The reported numbers can look compelling, but in ordinary usage I didn't see the same benefit. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.