Spent some time on this today and wanted to compare notes.

I've gone back and forth between vLLM and llama.cpp for Llama. The interesting difference for me has been token cost.

Trying to avoid rebuilding the whole thing twice.

Posting this in case it saves someone else a little time.

I'm not trying to build a perfect architecture diagram. I just want a sane starting point around MLX and Threadripper box.

I'm especially interested in mistakes worth avoiding.

This is mostly for local work rather than public serving.

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

I tried this with DeepSeek using MLX on A6000. The headline result was better than I expected, but router policy still ended up being the limiting factor.

Happy to post more details if anyone wants them.

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

It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with workstation thermals.

Maybe obvious to some of you, but it clicked for me this week.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.

Has anyone else run into this?

I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling prefix caching.

Would appreciate practical advice rather than theory.

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

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

Short version: Ollama behaved well for me, but only after I stopped ignoring quant drift.

Curious if anyone else landed in the same place.

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

The more I work with local setups, the more it seems like workstation thermals decides everything downstream.

Wondering if this is just where the whole space is heading now.

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

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

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

I expected the biggest difference to come from the model, but the stack around it mattered at least as much.

Spent some time on this today and wanted to compare notes.

I've gone back and forth between SGLang and Ollama for Gemma. The interesting difference for me has been workstation thermals.

Trying to avoid rebuilding the whole thing twice.

Spent some time on this today and wanted to compare notes.

Two paths both seem viable, but they break down differently once KV cache pressure becomes important.

Interested in tradeoffs more than fan loyalty.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.

The part that surprised me was how much the boring details mattered more than the model choice.