+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: LLM Models (https://www.datasculpting.ai/inference/forumdisplay.php?fid=12)
+--- Forum: Open Weight (https://www.datasculpting.ai/inference/forumdisplay.php?fid=49)
+---- Forum: Recipes & Configurations (https://www.datasculpting.ai/inference/forumdisplay.php?fid=52)
+---- Thread: model fallback rules that work (/showthread.php?tid=615)
model fallback rules that work - cozyarch - 09-30-2026
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 KV cache pressure.
Curious whether that lines up with what other people are seeing.
This is mostly for local work rather than public serving.
Re: model fallback rules that work - routerunner - 10-02-2026
I don't think that's the whole story. I get why that seems right, but on 5090 the bottleneck felt elsewhere. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.
Re: model fallback rules that work - torchkite - 10-04-2026
Quote:I don't think that's the whole story. I get why that seems right, but on 5090 the bottleneck felt elsewhere. Not saying you're wrong, just that I woul
Good point. The pattern around long context has been pretty similar for me. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.
Re: model fallback rules that work - rocmriley - 10-04-2026
Thanks, that helps. The pattern around VRAM headroom has been pretty similar for me. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
Re: model fallback rules that work - mlxgarden - 10-04-2026
Ran a quick check on my box. On A6000 with Ollama, I saw basically the same behavior.
Re: model fallback rules that work - llm_wren - 10-04-2026
That matches what I was seeing. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.
Re: model fallback rules that work - dockkite69 - 10-04-2026
A couple questions before I guess too hard. Is this local-only or are you routing anything externally?
Re: model fallback rules that work - torchshed - 10-04-2026
What worked for me was pretty simple. On dual 3090, the easiest win is often to back off the aggressive settings and see whether the problem survives. Once you know the plain path works, then add the faster path back in piece by piece.
Re: model fallback rules that work - fastbench - 10-06-2026
A couple questions before I guess too hard. What model, what runtime, and roughly what context size?