Reflection AI Unveils Beam Model
Reflection AI has officially unveiled Beam, its first frontier, open-weight AI model. The Brooklyn-based startup, founded two years ago, claims Beam matches the performance of leading Chinese open models on advanced reasoning benchmarks at significantly lower costs. This development could intensify the competition to build a Western alternative to models like DeepSeek, Qwen, and Z.ai.
The company provided new details in a blog post, describing Beam as a text-only mixture-of-experts model. It was trained using high-compute reinforcement learning to excel at reasoning, coding, and agentic tasks, reportedly at a fraction of the token cost and inference time compute compared to its rivals.
Technical Specifications and Performance
Beam is a 501-billion-parameter model with 23 billion active parameters. It underwent pretraining on 23.8 trillion tokens and features a 1 million token context window. For comparison, Z.ai’s GLM-5.2 has approximately 744 billion total parameters with 40 billion active parameters.
While Reflection’s performance claims await independent verification, the company states that Beam scores on par with Z.ai’s GLM-5.2 on advanced reasoning benchmarks and surpasses current leading Western open models, all while utilizing three to four times less inference compute. Reflection describes Beam as a 'workhorse model' suitable for enterprises, the public sector, and developers.
Market Positioning and Competition
Reflection is positioning itself against established closed labs like Anthropic and OpenAI, as well as popular open models from Chinese developers, and Western players such as Mistral, Meta, and Cohere. Its closest U.S. rival might be Inkling, an open model from Mira Murati’s Thinking Machines Lab. Reflection’s benchmarks indicate Beam outperforms Inkling on four coding tests where both models reported results, though Inkling is a multimodal model while Beam is text-only.
Company Background and Funding
Founded in 2024 by two former Google DeepMind researchers, Reflection has secured approximately $4.7 billion in funding from investors including Nvidia, Sequoia Capital, and Lightspeed Venture Partners. Its most recent funding round valued the company at a $25 billion pre-money valuation.
The startup has also focused on securing compute resources, a critical component for training frontier models that can attract customers away from both closed models and more affordable Chinese open-weight models. This summer, Reflection finalized deals totaling over $7 billion with SpaceX and Nebius to gain access to Nvidia’s GB300 chips through 2029.
The 'AI Factory' Vision
Reflection aims to target enterprises and sovereign nations with Beam and its future models. The core proposition involves building 'AI factories,' a product enabling institutions to create their own customized, local AI systems by training Reflection’s models on their proprietary data. Nvidia CEO Jensen Huang, whose company is an investor, has been a vocal proponent of the 'AI factory' concept and advocates for a robust open AI ecosystem, which would also benefit Nvidia’s GPU business.
Hedge funds and trading firms are reportedly keen on developing such systems. Reflection has already initiated testing for a sovereign AI factory partnership with Shinsegae Group in South Korea.
Reflection plans to release Beam’s weights and comprehensive technical details this month. Distribution will occur through hyperscalers and neoclouds, with integrations across open-source libraries at launch.
Reflection did not respond to requests for additional information.
