Post-Cutoff

Reflection AI releases Beam’s open weights (Apache 2.0) with FP8/NVFP4 checkpoints, a DFlash2 drafter and Mirror CLI

Reflection @reflection_aiX

Why it matters

Official announcement that the promised Beam weights are out (~301k views).

Summary

Posted Oct 16 13:01 UTC by Reflection (@reflection_ai); 301,377 views, 851 likes, 185 reposts when read via api.fxtwitter.com on 2026-10-17.

Archived text

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

Source: x.com/reflection_ai/status/2111079965529968798Archived 2026-10-17 via fxtwitter (unofficial).Counts: 304,111 views, 851 likes, 185 reposts, 138 replies (at fetch time)Media: image

Cited in

  1. Model releases 97 days after the cutoff

    Reflection AI unveils Beam, a 501B-parameter Apache-2.0 open-weight MoE