Loading timeline…
20252020s
Liang Wenfeng
Founder of DeepSeek, whose R1 model showed frontier reasoning could be trained and released openly at a fraction of the expected cost.
Organizations
DeepSeekHigh-Flyer
Major Achievements
- •Founded DeepSeek in 2023, funded by the quantitative hedge fund High-Flyer that he had co-founded.
- •Released DeepSeek-R1 in January 2025 with open weights, matching leading reasoning models at far lower reported training cost.
- •The release forced a public reassessment of how much capital a frontier model actually requires.
- •Made open-weight releases central to China's position in the frontier model race.
Key Papers
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
DeepSeek-AI · arXiv · 2025
- DeepSeek-V3 Technical Report
DeepSeek-AI · arXiv · 2024