GB300

Headlines · 2

1. 4d
   vLLM adds day-0 support for Reflection's 501B Beam model with Blackwell optimizations

   vLLM published day-0 serving support for Reflection's Beam, a 501B-parameter mixture-of-experts model with 23B active parameters per token, optimized for NVIDIA Blackwell. The engine handles reasoning, tool calling and structured outputs out of the box and can run Beam with tensor, data or expert parallelism, with a Docker image and launch command provided. The post says coding agents autonomously tuned the implementation so batch-1 decode is 2.71× faster and prefill 1.87× faster than the first version, and that Beam's DFlash2 drafter reaches 708 tokens/s for a single request on TP8 with an acceptance length of 3.7. It also describes day-0 reinforcement-learning support built with the SkyRL team, reporting AIME25 average accuracy rising from about 61% to about 79% after 50 training steps.

   vLLM Blog · 🔥 0
2. 5d
   Qwen3.8-Flash-Next hits 685 output tokens/s per user in TokenSpeed test

   Qwen's official X account congratulated the LightSeek Foundation on TokenSpeed results, saying Qwen3.8-Flash-Next reached 685 output tokens per second per user on NVIDIA GPUs. LightSeek's description says the run used TokenSpeed on four NVIDIA GB300 GPUs with NVFP4, and it links a blog post detailing the kernel and serving optimizations behind the figure. Qwen added that it hopes the community will unlock more of its inference potential; the number comes from LightSeek's own benchmark rather than an independent measurement.

   Qwen (X) · 🔥 0

Experience and discussion from the community

Share my GB300 experienceAsk about GB300