+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Optimization & Performance (https://www.datasculpting.ai/inference/forumdisplay.php?fid=10)
+--- Forum: Context / KV Cache (https://www.datasculpting.ai/inference/forumdisplay.php?fid=41)
+--- Thread: draft model choices: setup advice (/showthread.php?tid=822)
draft model choices: setup advice - cacheline - 10-06-2026
Did a bit of testing and thought I'd post the result here.
I tried this with FLUX using TensorRT-LLM on dual 3090. The headline result was better than I expected, but structured output still ended up being the limiting factor.
Happy to post more details if anyone wants them.
My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.
Re: draft model choices: setup advice - cacheline - 10-07-2026
worth checking the driver first.
Re: draft model choices: setup advice - cacheline - 10-10-2026
Appreciate the detail. 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.
Re: draft model choices: setup advice - turboyard94 - 10-11-2026
Tried to reproduce it just now. On B200 with MLX, I saw basically the same behavior. I expected the model choice to matter more than it did.
Re: draft model choices: setup advice - llmarch51 - 10-11-2026
Interesting, I hadn't tried that. I ended up in about the same place using Ollama.
Re: draft model choices: setup advice - mangohouse - 10-11-2026
Good point. That matches what I saw on GB200. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.
Re: draft model choices: setup advice - cacheline - 10-11-2026
What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at batch size. Once you know the plain path works, then add the faster path back in piece by piece.