+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Generative AI (https://www.datasculpting.ai/inference/forumdisplay.php?fid=17)
+--- Forum: Video (https://www.datasculpting.ai/inference/forumdisplay.php?fid=70)
+--- Thread: vision-language models on consumer gear: worth it? (/showthread.php?tid=628)
vision-language models on consumer gear: worth it? - marbletrack - 09-30-2026
Wanted to compare notes with people doing similar things.
My current setup is PyTorch with mini PC, and I keep running into tokens/sec.
If you've tested something similar, what did you keep and what did you stop bothering with?
My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.
Re: vision-language models on consumer gear: worth it? - turboarch95 - 09-30-2026
Thanks, that helps. 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: vision-language models on consumer gear: worth it? - cachelab - 09-30-2026
Quote:Thanks, that helps. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest of t
That matches what I was seeing. The pattern around prefix caching has been pretty similar for me.
Re: vision-language models on consumer gear: worth it? - audiopoint18 - 09-30-2026
I'd approach it in a boring order. If you're on PyTorch, 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.
Re: vision-language models on consumer gear: worth it? - quietarch - 09-30-2026
following this because I'm in the same boat.
Re: vision-language models on consumer gear: worth it? - turboarch95 - 10-01-2026
Thanks, that helps. That matches what I saw on A6000. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.
Re: vision-language models on consumer gear: worth it? - cachelab - 10-01-2026
Interesting, I hadn't tried that. I ended up in about the same place using Ollama.
Re: vision-language models on consumer gear: worth it? - quietarch - 10-02-2026
Interesting, I hadn't tried that. I don't think your conclusion is far off.
Re: vision-language models on consumer gear: worth it? - thetanest - 10-03-2026
Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher.