Reflection AI open-sources 501B MoE model Beam with 1M context, vLLM Day-0 support
woosuk_k · x · 2026-10-16
Reflection AI released Reflection-Beam-1.0 under Apache 2.0 with open weights, built with 40+ ecosystem partners and Day-0 support in vLLM.
Key points:
- Specs: 501B total parameters, 23B active per token, 1M-token context; combines hybrid attention, Harmony-format reasoning, and tool use
- Releases: BF16, FP8, and NVFP4 weights plus a DFlash2 drafter; weights and model card on Hugging Face
- Training tooling: the open-source harness used to train Beam, "Mirror CLI"
- RL: Day-0 support in SkyRL with LoRA, BF16 full fine-tuning, and MXFP8 RL, using vLLM rollouts
- Serving: vLLM team co-optimized the path from fused kernels to DFlash
Related event: Reflection Open-Sources Beam: 501B-Parameter MoE with 1M Context(21 posts)→
More from Infra
- Neocloud stocks sink on Gartner's "80% will fail" report; analyst calls it garbage — McDonaghMatthew · 2026-10-16
- JapanFold Claims One Tenstorrent Server Beats an Nvidia DGX B200 on Bio Models at a Third of the Price — DavidBennett__ · 2026-10-16
- vLLM ships Day-0 RL support for Beam in SkyRL, +62% LoRA decode throughput — vllm_project · 2026-10-16
- DFlash2 speeds Beam decode up to 3.03× across benchmarks, vLLM reports — vllm_project · 2026-10-16
- Inside Beam: 501B params, only 23B active per token via 8-of-256 expert routing — vllm_project · 2026-10-16
- CoreWeave founders cashed out $3.3 billion before the stock's meltdown — SumitGup · 2026-10-16