Reflection AI unveils Beam, an open-weight model to challenge China

Reflection AI has unveiled Beam, an open-weight artificial intelligence model designed to match the performance of leading Chinese systems while running on a fraction of the computing power.

The two-year-old startup, founded in 2024 by two former Google DeepMind researchers, is positioning the text-only system as a direct Western alternative to open offerings from Chinese developers such as DeepSeek, Qwen, and Z.ai, as well as closed models from OpenAI and Anthropic. Reflection said it plans to release Beam’s weights and technical documentation later this month across hyperscalers, neoclouds, and open-source libraries.

How Beam compares on compute and scale

Beam is built on a mixture-of-experts architecture with 501 billion total parameters, of which 23 billion are active during any single pass. Pre-trained on 23.8 trillion tokens, the model supports a context window of one million tokens.

By comparison, Z.ai’s GLM-5.2 runs on approximately 744 billion parameters, activating 40 billion. Reflection claims Beam matches GLM-5.2 on advanced reasoning benchmarks while using between three and four times less inference compute. Those benchmark claims have not been independently audited.

Against domestic competition, Reflection said Beam outscored Inkling—the multimodal open model released in July by Mira Murati’s Thinking Machines Lab—on four coding tests where comparable figures exist, though Beam remains strictly text-only.

The enterprise push and sovereign AI factories

Reflection has raised roughly $4.7 billion to date from investors including Nvidia, Sequoia Capital, and Lightspeed Venture Partners, reaching a $25 billion pre-money valuation in its latest funding round, according to PitchBook data.

Alongside model development, the company has spent aggressively to lock down physical infrastructure. Over the summer, Reflection agreed contracts worth more than $7 billion with SpaceX and Nebius to secure supply of Nvidia’s GB300 processors through 2029.

The commercial strategy centres on what the company calls "AI factories"—dedicated environments that allow private companies and sovereign governments to fine-tune Reflection’s open weights on proprietary data without leaking internal information. Reflection has already started testing a sovereign deployment in South Korea with retail conglomerate Shinsegae Group, while hedge funds and proprietary trading houses are also examining the setup.