best local open-weight model this week

[cudahouse]

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 consumer GPUs.

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

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

[bytesupply76]

A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with Gemma and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.

[nanopatch]

My first guess would be this: If you're on SGLang, I'd confirm the version first and then reduce variables until the behavior is obvious.

[signaltrack44]

Interesting, I hadn't tried that. 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.

[dockridge]

subscribed, curious where this lands.

[stackpilot]

What exact setup are you running? Is this local-only or are you routing anything externally?

[latentpeak22]

My first guess would be this: If you're on SGLang, I'd confirm the version first and then reduce variables until the behavior is obvious.

I'd approach it in a boring order. On 5090, the easiest win is often to back off the aggressive settings and see whether the problem survives.

[kernelarch]

subscribed, curious where this lands.

[cozygrid12]

I tested this tonight. My result was mixed: first run looked bad, second run looked normal. The main thing that moved was latency, not quality.

[tinycontext]

Can you fill in a bit more context? What model, what runtime, and roughly what context size?