best local open-weight model this week with TensorRT-LLM

[vramforge]

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

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

Curious if anyone else landed in the same place.

[cudapoint]

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. Once you know the plain path works, then add the faster path back in piece by piece.

[kernelgrid]

I'd start with the simpler stack.

[turboarch95]

Appreciate the detail. I ended up in about the same place using Ollama. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[kernelhouse70]

I'd approach it in a boring order. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at VRAM headroom. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[thetaplane]

Can you fill in a bit more context? Are you testing this on 5090 or something else?

[northgarden37]

Can you fill in a bit more context? Are you testing this on 5090 or something else?

I'd approach it in a boring order. If you're on LM Studio, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[routecraft]

Tried to reproduce it just now. On Threadripper box with vLLM, I saw basically the same behavior. I expected the model choice to matter more than it did.

[torchtrack24]

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

[torchsupply99]

Need a little more detail here. Do you see the same thing with a smaller batch or shorter prompt?