Skip to content
→ All articles

AI

Reflection AI Unveils Beam, Its First Open-Weight Model With 501 Billion Parameters to Rival Chinese Models

Reflection AI Unveils Beam, Its First Open-Weight Model With 501 Billion Parameters to Rival Chinese Models

On October 5, 2026, Reflection AI announced Beam, the company's first frontier open-weight AI model. Beam is a mixture-of-experts model with 501 billion total parameters, of which only 23 billion are active per request, pretrained on 23.8 trillion tokens with a context window of up to one million tokens. The company describes Beam as built on compute-intensive reinforcement learning aimed at reasoning, coding, and agentic performance.

In benchmarks the company published, Beam scores 80.9 on SWE-Bench Verified, 80.1 on Terminal-Bench v2.1, and 97.8 on AIME 2026. Reflection AI says these results match Z.ai's GLM-5.2 (a Chinese open model with roughly 744 billion total parameters and 40 billion active) and beat today's leading Western open models, while using three to four times less inference compute, according to the company.

Full Beam weights will be released under the open Apache 2.0 license later in October 2026, Reflection AI said.

Source: TechCrunch

Share WhatsApp X Telegram

Need filming at the level you just read about? I reply within 24 hours.

Start your project