Beta The Briev beta is out. Free on iPhone via TestFlight — install it in under a minute.

Join the beta ↗
Briev
Live
Technology

Reflection AI launches Beam, a low-cost open-source model aimed at Western users

Reflection AI, a U.S. startup founded by ex-DeepMind researchers, released its first open-weight model called Beam, which it says outperforms other Western open models while using far less compute.

Reflection AI, a startup created in 2024 by former DeepMind staff, announced the debut of its open-weight model Beam, targeting businesses and governments that want sovereign AI without relying on Chinese code. Beam is marketed as a high-performing workhorse for coding and agentic applications, requiring only a fraction of the compute power needed by similar open models, which translates to lower operating costs. According to co-founder and CEO Misha Laskin, the model’s capabilities are comparable to China’s GLM-5.2 and are closing the gap with Alibaba’s Qwen 3.8-Max, though it remains behind the leading closed models from OpenAI and Anthropic.

The release comes as Western firms increasingly adopt Chinese open-source AI to cut expenses, even as U.S. and U.K. security reviews warn that such models can aid hackers. Backed by investors including Nvidia, Sequoia and Citigroup, Reflection aims to provide an alternative for entities seeking control over AI deployment while contributing to regulatory discussions in Washington and London.

Why it matters

Beam offers a cheaper, locally controlled AI option for Western users, challenging reliance on Chinese open-source models and influencing future AI regulation.

In this story

open-source AIBeam modelReflection AIcoding taskssovereign AIAI regulationChinese modelssecurity concerns
Get the beta ↗