+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Optimization & Performance (https://www.datasculpting.ai/inference/forumdisplay.php?fid=10)
+--- Forum: Quantization (https://www.datasculpting.ai/inference/forumdisplay.php?fid=40)
+--- Thread: best small quant that still feels smart (/showthread.php?tid=211)
Pages:
1
2
Re: best small quant that still feels smart - cacheworks - 09-14-2026
Thanks, that helps. That matches what I saw on Threadripper box. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
Re: best small quant that still feels smart - vrambandit - 09-14-2026
What exact setup are you running? Are you testing this on 5090 or something else?
Re: best small quant that still feels smart - turboscope - 09-14-2026
Thanks, that helps. That matches what I saw on 5090. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.
Re: best small quant that still feels smart - southtinker - 09-14-2026
What exact setup are you running? What model, what runtime, and roughly what context size?
Re: best small quant that still feels smart - vrambandit - 09-15-2026
Quote:What exact setup are you running? What model, what runtime, and roughly what context size?
I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with Mixtral and compare from there. Not glamorous advice, but repeatable baselines save a lot of time.
Re: best small quant that still feels smart - nanobrook - 09-15-2026
Quote:Quote:What exact setup are you running? What model, what runtime, and roughly what context size?
Need a little more detail here. Is this local-only or are you routing anything externally?