← Back
DeepSeek R1
An open-weight Chinese AI model released by DeepSeek that surprised the industry with near-state-of-the-art performance at allegedly much lower cost.
2
sources
Mentioned by
All mentions
"This happened about a year ago in January 2025, when the open-weight Chinese company DeepSeek released DeepSeek R1, that I think it's fair to say surprised everyone with near-state-of-the-art performance, with allegedly much less compute for much cheaper."
From:
State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI | Lex Fridman Podcast #490
•
▶ 2:05
•
Jan 2026
Attribution: Lex describes the DeepSeek moment and how R1 surprised the industry
"And then DeepSeek are the people that did the training breakthrough, which is, they scaled the reinforcement learning."
From:
State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI | Lex Fridman Podcast #490
•
▶ 1:38:59
•
Jan 2026
Attribution: Nathan credits DeepSeek with the breakthrough in scaling reinforcement learning through their R1 model