Reflection AI has unveiled Beam, an open-weight AI model with 501 billion total parameters and 23 billion active per token. The US startup, founded in March 2024 by former Google DeepMind researchers, positions its model against Chinese rivals Z.ai’s GLM-5.2 and Alibaba’s Qwen 3.8-Max. Beam scored 80.9 on SWE-Bench Verified and 80.1 on Terminal Bench v2.1 while using 3 to 4 times less inference compute. The company has raised over $4 billion and carries a $25 billion pre-money valuation, backed by Nvidia, Sequoia and Citigroup. Full model weights and technical documentation are slated for release in October 2026 under the Apache 2.0 license.
Source: Read the original article

