{"version":1,"type":"story","url":"https://digestai.news/story/reflection-unveils-beam-an-open-model-it-says-rivals-glm-5-2-on-reason","json":"https://digestai.news/story/reflection-unveils-beam-an-open-model-it-says-rivals-glm-5-2-on-reason.json","markdown":"https://digestai.news/story/reflection-unveils-beam-an-open-model-it-says-rivals-glm-5-2-on-reason.md","slug":"reflection-unveils-beam-an-open-model-it-says-rivals-glm-5-2-on-reason","headline":"Reflection AI releases Beam, a 501B open-weight model claiming parity with Chinese rivals at lower compute cost","summary":"Reflection AI, a Brooklyn-based startup founded by former DeepMind researchers, has unveiled Beam, its first open-weight model. The 501-billion-parameter Mixture-of-Experts model activates 23 billion parameters per token and targets coding, reasoning, and agentic workloads. Reflection claims Beam matches Z.ai's GLM-5.2 on advanced reasoning benchmarks while using three to four times less inference compute, and approaches Alibaba's Qwen 3.8-Max on coding tasks. The model was pretrained on 23.8 trillion tokens in under four weeks using 6,144 NVIDIA GB300 GPUs, followed by a large-scale reinforcement learning run generating over 100 million rollouts on 10,500 GPUs. Beam is currently in final red-teaming with early access via a waitlist; Apache 2.0 weights are scheduled for release later this month.\n\nReporting differs on funding details. Semafor and TechCrunch cite a $25 billion pre-money valuation, but TechCrunch puts total capital raised at roughly $4.7 billion from backers including Nvidia, Sequoia Capital, and Lightspeed Venture Partners, while Semafor lists investors as Nvidia, Sequoia, and Citigroup. Only TechCrunch and Unite.ai report Reflection's compute deals worth over $7 billion with SpaceX and Nebius for Nvidia GB300 chips through 2029. Unite.ai uniquely details Reflection's sovereign AI cloud partnership with Shinsegae Group in South Korea and its certification with the U.S. Department of Energy's Genesis Mission. Marktechpost provides the most granular technical specifications, including architecture details like SandwichNorm and FP32 residual accumulation, and reports a 92.3% training goodput with nine semi-automatic rewinds.","keyPoints":["Beam is a 501B MoE model with 23B active parameters, scoring 80.9 on SWE-bench Verified and 80.1 on Terminal Bench v2.1 per Reflection's benchmarks.","Reflection claims Beam matches GLM-5.2 on reasoning using 3-4x less inference compute; Apache 2.0 weights coming later this month.","Sources differ on investors: TechCrunch cites Lightspeed and $4.7B raised; Semafor cites Citigroup and $25B valuation."],"whyItMatters":"Beam represents a major Western open-weight entrant claiming Chinese-model parity at lower compute cost, targeting enterprises and sovereign nations seeking non-Chinese AI infrastructure. Its release intensifies competition in the open-model layer and could influence U.S. AI regulation debates.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Reflection AI","Z.ai","Alibaba","Nvidia","Sequoia Capital","SpaceX"],"models":["Beam","GLM-5.2","Qwen 3.8-Max","Kimi K3","Nemotron 3 Ultra","Inkling"],"people":["Misha Laskin","Jensen Huang","Mira Murati"]},"firstPublishedAt":"2026-10-05T19:25:01Z","updatedAt":"2026-10-05T21:04:20Z","sourceCount":4,"hasPrimarySource":false,"sources":[{"outlet":"semafor.com","title":"Reflection unveils Beam, an open model it says rivals GLM 5.2 on reasoning while using 3x-4× less compute and approaches Qwen3.8-Max on coding and agentic tasks","url":"https://semafor.com/article/10/05/2026/reflection-ai-unveils-an-open-source-answer-to-chinese-labs","publishedAt":"2026-10-05T19:25:01Z","type":"press","primary":false,"lead":true},{"outlet":"MarkTechPost","title":"Reflection AI Introduces Beam: A 501B Open-Weight MoE Model With 23B Active Parameters for Coding and Agentic Workloads","url":"https://marktechpost.com/2026/10/05/reflection-ai-introduces-beam-a-501b-open-weight-moe-model-with-23b-active-parameters-for-coding-and-agentic-workloads","publishedAt":"2026-10-05T21:04:20Z","type":"press","primary":false,"lead":false},{"outlet":"Unite.AI","title":"Reflection AI Unveils Beam, a 501B-Parameter Open-Weight Model","url":"https://unite.ai/reflection-ai-unveils-beam-a-501b-parameter-open-weight-model","publishedAt":"2026-10-05T19:46:03Z","type":"press","primary":false,"lead":false},{"outlet":"TechCrunch AI","title":"Reflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute cost","url":"https://techcrunch.com/2026/10/05/reflection-debuts-beam-a-open-weight-ai-model-to-rival-chinese-models-at-lower-compute-cost","publishedAt":"2026-10-05T19:33:53Z","type":"press","primary":false,"lead":false}],"sourceNotes":{"agree":"All four articles report Reflection AI launched Beam, a 501B-parameter open-weight MoE model with 23B active parameters, claiming performance rivaling Chinese models GLM-5.2 and Qwen 3.8-Max at lower inference compute.","differ":["TechCrunch reports $4.7B total raised from Nvidia, Sequoia Capital, Lightspeed; Semafor lists Nvidia, Sequoia, Citigroup at $25B valuation.","Only TechCrunch and Unite.ai mention $7B+ compute deals with SpaceX and Nebius for GB300 chips through 2029.","Only Unite.ai details Shinsegae sovereign AI cloud partnership and DOE Genesis Mission certification."]},"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Reflection AI releases Beam, a 501B open-weight model claiming parity with Chinese rivals at lower compute cost\", 5 October 2026, https://digestai.news/story/reflection-unveils-beam-an-open-model-it-says-rivals-glm-5-2-on-reason","publisher":"Digest AI","title":"Reflection AI releases Beam, a 501B open-weight model claiming parity with Chinese rivals at lower compute cost","datePublished":"2026-10-05T19:25:01Z","url":"https://digestai.news/story/reflection-unveils-beam-an-open-model-it-says-rivals-glm-5-2-on-reason"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}