Announced 5 Oct 2026 · Sources checked
What did Reflection announce?
Beam is a text-only sparse mixture-of-experts model built for coding, reasoning, tool use, and agentic tasks. Reflection says it pretrained the model on 23.8 trillion tokens and ran reinforcement learning on 10,500 NVIDIA GB300 GPUs for four weeks.
How strong is the evidence?
Reflection published benchmark tables and methodology notes, but the results are company-reported. The company says Beam is competitive with GLM 5.2 on some coding and agentic tests while larger open models remain ahead on several measures. AiLookout has not independently tested Beam.
Can developers use Beam now?
Only a selected early-access group can use the preview. Reflection says it plans to release the model weights under Apache 2.0 later in October, together with a model card, technical report, documentation, and developer tools.
Common questions
Are Beam’s weights available now?
No. Reflection says the weights will be released later in October 2026 after final evaluations.
Is Beam fully open source?
Not yet. Reflection calls it open-weight and promises an Apache 2.0 release, but readers should review the final license and published artifacts when they arrive.
What to remember
Treat Beam as a limited preview until the weights, model card, and technical report are public.
Sources & further reading
How this story was made
Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. Cover photography is illustrative and does not depict the specific announcement or product.
Our editorial standards
