Beam Proves RL Now Eats More Compute Than Pre-Training
ReflectionAI CEO Misha Laskin explains why training Beam required 10,000 GB300 GPUs on reinforcement learning vs 6,000 on pre-training.
Beam: The Great American Open Model with ReflectionAI Co-Founder and CEO Misha Laskin
ReflectionAI co-founder and CEO Misha Laskin joins Sarah Guo and Elad Gil to discuss the launch of Beam, a 500-billion-parameter open-weight reasoning model. Laskin outlines why open models will capture the vast majority of enterprise token demand through an 'ownership' rather than 'rental' compute model, the technical requirements of scaling reinforcement learning, and how open source AI is critical for cybersecurity and geopolitical resilience.
Key takeaways
- ReflectionAI trained Beam, a 500-billion-parameter model with 23 billion active parameters, spending more FLOPs on reinforcement learning than on pre-training. Read more →
- Elad Gil views Chinese open-source AI as a direct subsidy to Western enterprises, giving builders frontier intelligence without paying the research bills. Read more →
- When enterprise spending on closed-model APIs crosses $100 million annually, the economic math flips from renting proprietary tokens to running private, owned infrastructure. Read more →
- ReflectionAI launched Beam, a 500-billion-parameter open-weight reasoning model, targeting an ownership rather than rental compute model for enterprises. Read more →
- ReflectionAI released Beam, a 500-billion-parameter open-weight reasoning model built on the belief that open architectures beat closed safety moats. Read more →
5 articles from this episode
Why Chinese Open-Source AI Is an Infrastructure Trap
Misha Laskin and Elad Gil debate Chinese open weights: are they free enterprise subsidies or Trojan horses for Huawei silicon?
Why Enterprise AI Will Shift From Renting to Owning Tokens
Misha Laskin explains why $100M API bills will push enterprise token volume to open-weight models like ReflectionAI's Beam.
Misha Laskin: How to Win When AI Margins Hit Zero
ReflectionAI CEO Misha Laskin explains why intelligence density, compute scale, and customer trust determine who survives AI commoditization.
Why Closed-Door AI Safety Fails Cyber Defense
ReflectionAI CEO Misha Laskin explains why open-weight models and Linus's Law protect infrastructure better than closed AI labs.
More No Priors episodes
- Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of FractileOct 4 · 5 articles
- Re-Founding Incumbents for the AI Era with Sequence Holdings Co-Founder and CEO Michael LeeSep 27 · 5 articles
- Coinbase’s Everything Exchange: Agentic Finance, Stablecoins & Tokenization with CEO Brian ArmstrongSep 13 · 6 articles
- Redefining Chip Architecture with Arm CEO Rene HaasSep 6 · 7 articles