Reflection AI 发布其首个开源权重模型 Beam，501B MoE 架构，权重以 Apache 2.0 许可开放，与 40+ 伙伴生态合作。发布内容包括权重、tokenizer、参考实现、BF16/FP8/NVFP4 权重、DFlash2 drafter，以及训练所用开源 harness Mirror CLI，模型卡在 Hugging Face 提供。NVIDIA AI 补充称该模型在 NVIDIA GB300 NVL72 上训练，在 Blackwell Ultra 上开箱即达 70K tokens/sec/GPU 以上，并提供 Beam 1.0 NVFP4 checkpoint 下载：https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4
同一新闻，精选展示《Reflection 发布开源模型 Beam：501B 参数 MoE，Apache 2.0 协议》
Welcome Beam!
Congrats to @ReflectionAI on launching their first open-weight model.
The 501B MoE model was trained on NVIDIA GB300 NVL72 and delivers over 70K tokens/sec/GPU on Blackwell Ultra out of the box on day 0.
🤗 Beam 1.0 NVFP4 checkpoint on @huggingface: https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4
Beam is now available across platforms, with open weights under the Apache 2.0 license - in collaboration with an ecosystem of 40+ partners. We're releasing: - Weights, tokenizer, and reference implementation - BF16, FP8 and NVFP4 weights and a DFlash2 drafter - The open source harness we built and trained Beam on: Mirror CLI Weights and model card on Hugging Face: https://huggingface.co/reflection在 X 查看被引用的帖子
来源：NVIDIA AI · x.com