Reflection AI unveils Beam, an open-weight model it says matches top Chinese rivals with 3–4x less compute

Reflection AI, a two-year-old US start-up, officially unveiled Beam, its first frontier open-weight model, on 5 October 2026, TechCrunch reported, confirming weekend reporting by Axios.

What Reflection says about Beam

• A text-only mixture-of-experts model with 501 billion parameters, of which 23 billion are active per token; pretrained on 23.8 trillion tokens, with a 1-million-token context window.
• Trained with high-compute reinforcement learning for reasoning, coding and agentic tasks.
• On advanced reasoning benchmarks it scores on par with Z.ai’s GLM-5.2 (about 744 billion parameters, 40 billion active) and beats leading Western open models while using “3-4x less inference compute”, the company says.
• Reflection’s own charts show Beam ahead of Thinking Machines Lab’s Inkling on four coding tests, though Inkling is multimodal and Beam is not.

Important: these are the company’s own benchmark claims; TechCrunch notes they have not been independently verified. Reflection says the weights and full technical details will be released this month.

The company

Founded in 2024 by two former Google DeepMind researchers, Reflection has raised roughly $4.7 billion (per PitchBook) from backers including Nvidia, Sequoia and Lightspeed; its last round valued it at $25 billion pre-money. This summer it signed compute deals worth more than $7 billion with SpaceX and Nebius for Nvidia GB300 chips through 2029. It pitches Beam to enterprises and governments as the basis for customised local “AI factories”, and is testing a sovereign partnership with South Korea’s Shinsegae Group.