BLAST RADIUS

Open-weight & owned stack · Quantization efficiency

R9V Improves Qwen Flash Next Inference Performance

Sep 14, 2026

R9V added crash fixes, streaming diagnostics, pinned images, and Q4_K_XL support while reporting roughly 100 tokens per second on two R9700 GPUs.

Read the original at reddit.comOpens the publisher's site in a new tab

More in Open-weight & owned stack