Reflection: Beam is now available
blog
Fields
id
sources/reflection-beam-available-2026-10-16Publisher
Reflection
Kind
blog
Published
2026-10-16
Accessed
2026-10-18
notes
Fetched 2026-10-18. Availability post dated 2026-10-16. States the weights are released on Hugging Face under Apache 2.0 in BF16, FP8 and NVFP4, with a speculative-decoding drafter. Names day-zero inference support in Hugging Face Transformers, vLLM (Inferact), SGLang (RadixArk), Unsloth, Axolotl, TokenSpeed (LightSeek Foundation) and NVIDIA Dynamo. Also names LM Studio and Ollama among launch platforms. Does not document expert-offload behavior or a quantisation recipe beyond the three shipped precisions. Figure 2 caption says Reflection's production dashboard is powered by SGLang serving Beam on NVIDIA Blackwell.
Notes
Fetched 2026-10-18. Availability post dated 2026-10-16. States the weights are released on Hugging Face under Apache 2.0 in BF16, FP8 and NVFP4, with a speculative-decoding drafter. Names day-zero inference support in Hugging Face Transformers, vLLM (Inferact), SGLang (RadixArk), Unsloth, Axolotl, TokenSpeed (LightSeek Foundation) and NVIDIA Dynamo. Also names LM Studio and Ollama among launch platforms. Does not document expert-offload behavior or a quantisation recipe beyond the three shipped precisions. Figure 2 caption says Reflection's production dashboard is powered by SGLang serving Beam on NVIDIA Blackwell.
Cross-references
This record references no other records.