JustGhostIt — Frontier Models archive, page 9.
Reinforcement learning breakthroughs enable math reasoning surpassing DeepSeek without distilling R1. This 2025-2026 AI industry extension topic complements existing timelines.
At the Paris AI Action Summit, global leaders and industry reps clashed over open vs. closed models and safety vs. innovation, while France pledged massive private-sector AI investments.
DeepSeek-R1, an open-source reasoning model by DeepSeek, sparks global debate with low-cost training and MIT license. Its performance rivals OpenAI's o1, while its API pricing delivers a significant shock to the market.
OpenAI launches Operator, an agent that autonomously navigates browsers to book tickets and shop. This marks a major step in integrating vision, reasoning, and tool use.