Launches
Flare packages Qwen Flash Next for a single 96 GB Blackwell GPU
Community reports 4 sources
Q
Qwen3.8 Flash Next Flare
Model Oct 18, 2026
Local Inference Lab's alpha Qwen3.8 Flash Next package with prepared kernels and execution graphs for the Flare runtime on GB202 Blackwell hardware.
Visit primary link
2 sources attached
What happened lately
Qwen3.8 Flash Next Flare timeline
This community alpha distribution prepares Qwen3.8 Flash Next for Flare rather than standard inference engines. Its GB202 configuration targets a single RTX PRO 6000 Blackwell with 96 GB of VRAM, with kernels and execution graphs bundled alongside the model.
The guide keeps the 26.8 GiB embedding table on disk or in system RAM and budgets approximately 110 GB of disk and 64 GiB RAM for its RAM-backed examples. It does not establish an independently verified throughput result. The custom license restricts reuploads; it must not be described as Apache 2.0 alone.