Nvidia-backed Reflection unveils first AI model to take on Chinese open models
Beam is designed to run on Nvidia GPUs with 3 to 4 times less computing power than comparable open models, Reflection AI said.
- Nvidia-Backed Reflection AI debuted Beam, its first open-weight model, on Monday to compete with lower-cost Chinese AI systems like DeepSeek and Kimi in coding and agentic tasks. Founded by former DeepMind researchers, the startup positions Beam as a "workhorse" for Western organizations.
- CEO Misha Laskin noted that businesses and governments want to "own" and control their AI infrastructure without relying on Chinese models. The Nvidia-backed startup developed Beam to address demand for sovereign systems that maintain security while avoiding geopolitical dependencies.
- Beam contains 501 billion total parameters but activates only 23 billion per task, making it faster and cheaper to run. The model scores on par with China's Z.ai's GLM-5.2 on advanced reasoning benchmarks while using 3-4 times less inference compute.
- Western companies have recently embraced Chinese models to reduce costs for routine tasks, though security concerns regarding code vulnerabilities persist. Reflection's private alternative allows organizations to build customized AI systems while maintaining control over proprietary data.
- Reflection is already training its next model, which CEO Misha Laskin said would be "much more powerful and performant than Beam." The release positions the startup as America's answer to open-source Chinese AI, intensifying the competition for sovereign systems.
69 Articles
69 Articles
$25 Billion Reflection AI, Backed By Nvidia (NASDAQ: NVDA), Prepares To Launch Open-Source Model That Could Reshape Enterprise AI
Reflection AI, a New York City-based startup currently valued at $25 billion, has yet to ship a single product despite being founded roughly two and a half years ago. The company was founded by two former Google DeepMind researchers who were instrumental in building the foundational architecture of Google Gemini, the search giant’s flagship AI […]
Reflection AI's Beam Model Takes Aim at Chinese Open-Weight Lead with Efficiency Focus
Reflection AI launched Beam, a 501B-parameter open-weight MoE model with 23B active parameters. It claims parity with China's GLM-5.2 on reasoning benchmarks at 3-4x lower inference compute while excelling in coding and agentic tasks. Weights arrive later this month under Apache 2.0.
"I don't want an oligopoly": New open-weight AI models mount a comeback against China
A new crop of powerful, Western open-weight AI systems are mounting a comeback, with fresh models from Reflection and Mistral narrowing China's recent lead.Why it matters: These models and others will give businesses and governments alternatives to the closed ecosystems of OpenAI and Anthropic, as well as to Chinese options they may have eschewed for security reasons.But putting powerful model weights into the world also makes their safeguards e…
Reflection CEO Misha Laskin on launch of new AI model Beam, open-weight vs. closed model debate
Misha Laskin, Reflection co-founder and CEO, joins 'Squawk Box' to discuss the launch of its new open-weight AI model Beam, state of AI competition, open-weight vs. closed models, U.S.-China AI arms race, and more.
An artificial intelligence startup from the United States supported by Nvidia and the Trump government presented its first open source model, reinforcing US ambitions to compete with Chinese groups that dominate the market for public and customizable AI products. Exclusive material for subscribers. To have full access, access the link of the subject and register.
Coverage Details
Bias Distribution
- 40% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
