Reflection AI releases open-weight model Beam
The startup made Beam's weights and post-training recipes public, and its CEO Misha Laskin discussed the model's release and the company's valuation.
By Standard Topic. AI-generated from the reporting listed under Coverage. Published . Updated .
This story is no longer being updated. It reflects reporting up to .
What happened
Reflection AI, a startup founded by two DeepMind alumni, has released Beam, an open-weight AI model, and made its weights and post-training recipes publicly available. The company's CEO, Misha Laskin, discussed the release, the firm's valuation, and reports that Nvidia is expanding its stake in the company. Reflection AI is backed by Nvidia and aims to build a powerful open model that is cheaper than US rivals and safer than Chinese ones.
Why it matters
The reports describe Reflection AI as a £19bn threat to OpenAI, positioning its open model as a competitor to larger US AI companies. Laskin said the risks of artificial intelligence worry him.
Timeline
- Oct 16, 9:00 AM ET: Reports described two DeepMind alumni as having built a £19bn threat to OpenAI through Reflection AI, which is backed by Nvidia.
- Oct 16, 4:51 PM ET: Reflection AI CEO Misha Laskin discussed the release of Beam, the company's valuation, and reports of Nvidia expanding its stake on The Claman Countdown.
- Oct 16, 6:52 PM ET: Reflection AI open-sourced Beam's weights and post-training recipes.
Coverage
- Reflection AI open-sources Beam’s weights, post-training recipes (siliconangle.com)
- Reflection AI CEO confronts the risks of artificial intelligence: ‘It certainly worries me’ (foxbusiness.com)
- How two DeepMind alumni built a £19bn threat to OpenAI (Louisa Clarence-Smith)
More from Technology & AI
- Eight states sue OpenAI, alleging misleading AI safety claims
- OpenAI announces solution to Navier-Stokes math problem
- Datadog bets AI agents can handle testing of AI-written code
- Anthropic and OpenAI sketch wealth-sharing plans before IPOs
- Mamdani administration proposes AI customer service rules
- Florida engineers study how AI behaves in high-risk situations