Welcome to Issue #143 of One Minute AI, your daily AI news companion. This issue discusses a recent announcement from DeepSeek.
Introducing DeepSeek R1
DeepSeek, a Chinese AI research lab backed by High-Flyer Capital Management, has unveiled its latest reasoning models, DeepSeek-R1 and DeepSeek-R1-Zero, as open-source alternatives to proprietary systems like OpenAI's o1. Distributed under the MIT license, DeepSeek-R1 allows developers to freely use, modify, and commercialize the model. The model and its API, which supports fine-tuning and distillation, are accessible at chat.deepseek.com.
Alongside DeepSeek-R1, the lab released six distilled models ranging from 32 billion to 70 billion parameters, optimized for efficiency and claiming performance levels similar to OpenAI's o1-mini. These models are designed to tackle tasks in mathematics, code generation, and reasoning with competitive accuracy. Leveraging large-scale reinforcement learning in post-training, DeepSeek-R1 achieves high performance with minimal reliance on labeled data. The model scored 79.8% Pass@1 on the AIME 2024 benchmark, slightly surpassing OpenAI's o1-1217. This development follows DeepSeek's recent launch of DeepSeek-V3, touted as the best open-source model.
Want to help?
If you liked this issue, help spread the word and share One Minute AI with your peers and community.
You can also share feedback with us, as well as news from the AI world that you’d like to see featured by joining our chat on Substack.