Reflection AI unveils 501-billion-parameter Beam model to rival Chinese systems

Model architecture and benchmark claims

Reflection AI released Beam on Monday, introducing its first frontier open-weight artificial intelligence model designed to compete with Chinese systems in coding, reasoning, and agentic workflows. The text-only mixture-of-experts model contains 501 billion total parameters with 23 billion active parameters, having completed pre-training on 23.8 trillion tokens. Reflection built the system with a 1 million token context window, using high-compute reinforcement learning to reduce inference costs. According to the company, Beam operates with three to four times less inference compute than existing Western open models. Reflection claims the system matches the benchmark performance of Chinese startup Z.ai's GLM-5.2 model, which operates with approximately 744 billion total parameters and 40 billion active parameters.

Market shifts toward open-weight architectures

The release positions the startup against closed commercial providers, including OpenAI and Anthropic, alongside Chinese open-weight models from DeepSeek, Qwen, and Z.ai. Enterprise demand has increasingly turned toward customisable open-weight architectures to lower operational expenses and avoid closed subscription tiers. On cloud application platform Vercel, open-weight models processed 56% of AI Gateway tokens in August, rising from 7% in December. To power its systems, Reflection secured agreements worth more than $7 billion with SpaceX and Nebius to access Nvidia GB300 chips through 2029. Reflection also reported that Beam outscored Inkling, a multimodal open model released in July by Mira Murati's Thinking Machines Lab, across four shared coding evaluations.

Corporate valuation and strategic backing

Founded in 2024 by two former Google DeepMind researchers, Reflection attained a $25 billion pre-money valuation after raising roughly $4.7 billion from investors including Nvidia, Sequoia Capital, and Lightspeed Venture Partners. Nvidia alone contributed $800 million to the startup while providing direct chip access. Reflection has established deployment partnerships with the Pentagon and the US Department of Energy, in addition to agreements to build models for allies such as South Korea. The startup aims to present a domestic open-weight alternative as US officials evaluate security policies regarding Chinese software. Reflection co-founder and chief executive Misha Laskin described the strategic motivation during an interview in July.

Today the best open models are coming out of China. And as a result, global AI adoption is actually very much being built on top of Chinese models, including by American companies. And that's great, but we wish to build a thriving AI ecosystem that has a lot of competition and a lot of players, and we're building the counterbalance to that -- building great open models here in America.

Policy debate over open ecosystem

The debut of Beam coincides with active discussions in Washington regarding the role of foreign open-weight architectures in corporate infrastructure. US Treasury Secretary Scott Bessent floated potential sanctions on Chinese AI models over intellectual property concerns. In response, Nvidia led an industry letter arguing that an open-weight ecosystem remains necessary for domestic technological innovation. Nvidia chief executive Jensen Huang separately defended the corporate adoption of Chinese open models, citing their economic value for businesses. Reflection plans to market Beam as a foundation for enterprises and government agencies looking to build proprietary systems without relying on foreign weights or closed proprietary APIs.