

OpenAI Strawberry's Innards―How it works, and what comes next for OpenAI | Artificial Intelligence Masterclass
Sep 1, 2025
Dive into the workings of OpenAI Strawberry, where advancements in self-prompting techniques steal the spotlight. Explore the intriguing notion of post-labor economics, where machines take center stage, outperforming human capabilities. The discussion delves into the ethical frameworks guiding AI development, focusing on reducing suffering and increasing knowledge. Discover the balance between data insights and spontaneous thought processes, along with the competitive landscape shaping AI's future. Get ready for an enlightening journey into the next frontier of artificial intelligence!
AI Snips
Chapters
Transcript
Episode notes
Strawberry Uses Chain-of-Thought
- OpenAI Strawberry fine-tunes models specifically for chain-of-thought reasoning.
- This causes the model to self-prompt and think step-by-step rather than just answer directly.
Strawberry's Cost vs Benefit
- Strawberry is about 100 times more expensive than GPT-4 Mini for a modest performance improvement.
- The cost-to-benefit ratio makes it less practical for broad business use currently.
Multi-Agent Frameworks Potential
- Multi-agent frameworks can achieve similar or even better results at lower cost and with faster execution.
- These frameworks might outperform Strawberries' reasoning approach soon.