Model library / reflection
Beam is an open-source, instruction-tuned language model from Reflection using a sparse mixture-of-experts architecture.
Mixture of experts, count not published
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Beam is a 501-billion-parameter model built by Reflection. It is presented as an English-focused system for token-efficient reasoning, software-engineering tasks, agentic tool use and advanced scientific problem solving. The model is offered in BF16, FP8 and NVFP4 formats and is described as instruction-tuned.
In practice the model comprises 501.3 billion total parameters with 23 billion active parameters, a 1,048,576-token context window and an Apache-2.0 licence. The expert count is not disclosed and weight format is unpublished, so deployment must accommodate a very large parameter footprint and the specified licence terms.
Specification
| Repository | reflection/Beam-1.0-501B-A23B |
| Publisher | reflection |
| Published | 7 October 2026 |
| Architecture | BeamForCausalLM |
| Model type | beam |
| Parameters | 501,267,612,352 |
| Layers | 52 |
| Hidden size | 5,120 |
| Attention heads | 48, 8 key/value heads |
| Context length | 1,048,576 tokens |
| Vocabulary | 151,936 tokens |
| Licence | apache-2.0 |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 933.7 GiB | Full precision as released |
| fp8 | 466.8 GiB | 8-bit, near-lossless on most models |
| int4 | 233.4 GiB | 4-bit, smallest footprint |
The cache costs 208 KiB per token at 16-bit, so the full 1,048,576-token context of one request needs about 208.0 GiB. Concurrency multiplies that number, not the weights.
That figure is an upper bound: this model uses sliding-window attention with a 512-token window on some layers, so long requests cache less than this.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | more than 8 | — | — |
| RTX 5090 32GB | more than 8 | — | — |
| L40S 48GB | more than 8 | — | — |
| A100 80GB | more than 8 | — | — |
| H100 80GB | more than 8 | — | — |
| RTX PRO 6000 96GB | more than 8 | — | — |
| H200 141GB | 8 | 61.4 GiB | about 309,514 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve reflection/Beam-1.0-501B-A23B \ --tensor-parallel-size 8 \ --max-model-len 307200 \ --gpu-memory-utilization 0.90
Questions
The weights take about 934 GiB at the released precision. No single card in the table above holds that, so it needs several GPUs or a lower precision.
The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 1,048,576 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email hello@aiqre.com with the model name and roughly what volume you expect.
Also in the library
Ling-3.1-flash is a 572.1 billion-parameter hybrid reasoning language model with a 262,144-token context window.
IQuest-Q1 is a 320 billion-parameter text-only language model from IQuestLab aimed at agentic coding, reasoning and tool use.
MiMo-V2.6-Flash-MOPD is a 310.8 B parameter sparse-MoE model that merges domain-specialised teachers to reduce tool-call repetiti…
MiMo-V2.6-Pro-RL is a 524-billion-parameter sparse MoE model designed for reinforcement-learning-driven self-improvement across t…
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.