@TheAhmadOsman: Congrats to all the friends at Reflection We now have a SoTA American Opensource model
Summary
Congrats to Reflection AI on releasing Beam, a highly efficient agentic open model with 501B total parameters and 23B active, advancing the Western open-source frontier in coding and agentic tasks.
View Cached Full Text
Cached at: 10/06/26, 11:44 AM
Congrats to all the friends at Reflection
We now have a SoTA American Opensource model 🦅
Reflection (@reflection_ai): Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.
- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch
Full weights release this month.
Learn more about
Similar Articles
Western open-source AI Reflection finally shows up
Reflection AI, founded by former DeepMind researchers, released Beam, its first open-source model, positioning it as a Western alternative to Chinese open models with strong coding and agent capabilities and claimed 3-4x efficiency gains.
@woosuk_k: Congrats to @reflection_ai on the release! This time, we let AI agents drive model optimization autonomously, with perf…
Woosuk Kim congratulates Reflection AI on releasing Reflection-Beam-1.0, a 501B-total-parameter (23B active) MoE model with 1M-token context, hybrid attention, and Harmony-format reasoning, highlighting AI agents autonomously driving model optimization with tracked performance gains and Day-0 vLLM/SkyRL support.
Beam: Reflection's 501B open-weight model
Reflection AI introduced Beam, a 501B-parameter sparse MoE open-weight model (23B active) built for coding, reasoning, and agentic workloads, pretrained on 23.8T tokens and scaled with a high-compute RL run on 10.5K GB300 GPUs. Beam is positioned as advancing the Western open-weight frontier, approaching GLM 5.2 and Qwen 3.8-Max on coding/agentic tasks while using 3–4× less inference compute, with weights and a technical report due later this month.
Reflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute cost
Reflection AI 发布了其首个前沿开放权重模型 Beam——一个 5010 亿参数（230 亿活跃参数）的文本混合专家模型，宣称在先进推理基准上与 Z.ai 的 GLM-5.2 相当，但推理算力成本低 3-4 倍，旨在为西方提供对抗中国开源模型的低成本选择。
Beam
Reflection AI 推出 Beam：一款 5010 亿总参数（230 亿激活）的稀疏混合专家开源模型，面向编码、推理与智能体工作负载。