Beam announced as available across platforms with Apache 2.0 open weights
Reflection AI says it’s releasing the tokenizer and Mirror CLI, the open-source harness it used to train Beam.
TLDR
Reflection AI announced Beam’s cross-platform availability and open weights under Apache 2.0, citing collaboration with an ecosystem of 40-plus partners. It says it’s releasing a reference implementation, BF16, FP8 and NVFP4 weights, and a DFlash2 drafter. Hugging Face says the 501-billion-parameter text model is out on its platform, with a one-million-token context window.
Combined views
436.3K
9 Sources, first seen ago
