The A.I. Daily
Strata With Qwen3.8 Next Speeds Up Slow-Memory Rigs
A user reports that Strata with Qwen3.8 Next and IQ3_XXS weights reaches roughly 45-70 tokens/s on a DDR4 and 7900XTX setup, versus about 22.5t/s with tuned Llama CPP, with similar results on 12GB and 16GB GPUs.