Reflection AI Unveils Beam, Open-Weight Model Targeting Lower Compute Costs

Reflection shared details in a lengthy blog post Monday, confirming Axios reporting over the weekend that it was close to launch. Beam is a text-only mixture-of-experts model trained with high-compute reinforcement learning for reasoning, coding, and agentic tasks, according to the company. It has 501 billion total parameters, 23 billion active parameters, was pre-trained on 23.8 trillion tokens, and has a 1 million token context window.

Reflection says Beam performs on par with Z.ai’s GLM 5.2 on advanced reasoning benchmarks and outperforms today’s leading Western open models while using three to four times less inference compute. Z.ai’s GLM 5.2 has roughly 744 billion total parameters with 40 billion active, according to the comparison Reflection provided. The performance claims have not been independently verified.

Reflection calls Beam a “workhorse model” for enterprises, the public sector, and developers. The company’s benchmarks show Beam outscores Inkling, the open model from Mira Murati’s Thinking Machines Lab released in July, on four coding tests where both report results. Inkling is multimodal, while Beam is text-only, a difference Reflection disclosed in its comparison.

Reflection is aiming at closed labs including Anthropic and OpenAI, popular open models from Chinese developers, and Western open-model players such as Mistral, Meta, and Cohere. It was founded in 2024 by two former Google DeepMind researchers. The startup has raised roughly $4.7 billion from backers including Nvidia, Sequoia Capital, and Lightspeed Venture Partners, according to PitchBook, and its last round valued it at $25 billion pre-money.

The company has also secured compute. This summer, Reflection signed deals collectively worth more than $7 billion with SpaceX and Nebius to gain access to Nvidia’s GB300 chips through 2029. Compute access is central to training frontier models that can compete with closed models and cheaper open-weight models from Chinese labs.

Reflection is pitching Beam and future models to enterprises and sovereign nations through what it describes as “AI factories.” The product would let institutions build customized, local AI systems by training Reflection’s models on their own proprietary data. Nvidia CEO Jensen Huang, whose company backs Reflection, has championed the AI factory idea and pushed to strengthen the open AI ecosystem. Axios reported that hedge funds and trading firms are among those interested in building such systems. Reflection has begun testing a sovereign AI factory partnership with South Korea’s Shinsegae Group.

Reflection says it will release Beam’s weights and full technical details this month. Distribution will run through hyperscalers and neoclouds, with integrations across open source libraries at launch. Reflection did not respond in time to TechCrunch’s requests for more information.