student models for local use on M3 Max

[northquill]

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 token cost.

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

[torchtrack24]

A couple questions before I guess too hard. Are you testing this on B200 or something else?

[oxidemixer]

Tried to reproduce it just now. On B200 with LM Studio, I saw basically the same behavior. The main thing that moved was latency, not quality. I keep promising myself this is the last rebuild.

[cobaltlab]

I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with Mixtral and compare from there.

[amberpatch25]

Not sure I agree there. I get why that seems right, but on 4090 the bottleneck felt elsewhere. Worth testing the simpler explanation before adopting the more interesting one.

[cobaltlab]

A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with Kimi and compare from there.

[quantshed]

Interesting, I hadn't tried that. I don't think your conclusion is far off.

[prompttinker44]

My first guess would be this: If you're on vLLM, 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.

[quantshed]

Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher.

[dockember]

Ran a quick check on my box. My result was mixed: first run looked bad, second run looked normal.