Blog
Written versions of our YouTube videos, so you can check the numbers and tables at your own pace.
-
Strata on RTX 3090: 5 Quantizations of 125B Measured, IQ3_S Prefill 2045 to 207 at 32GB RAM (youtube.com)
We measured five quantizations of Qwen 3.8 Flash-Next 125B with Strata on one RTX 3090. Korean accuracy overlapped; at about 32GB RAM, IQ3_S prefill fell from 2045 to 207 tok/s.