Open-source DeepSeek R1 matches OpenAI’s o1 for just 3% to 5% of the cost. How so? This AI LLM taught itself completely through trial and error, or reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, similar to how humans learn.
RL is basically a method in which an LLM is rewarded for making good decisions and punished for making bad ones, without knowing which one is which. After a number of decisions, it learns to follow a path that was reinforced by those results. DeepSeek R1 mainly learns through mechanical reinforcement learning, or figuring things out by experimenting and getting feedback on what works best. Most importantly, it runs queries at a mere $0.14 per million tokens compared to OpenAI’s $7.50, making it roughly 98% cheaper. Try it out here.
- Transform your reality and do everything you love in totally new ways. Welcome to Meta Quest 3S. Now you can get the Batman: Arkham Shadow* and a...
- Explore thousands of unreal experiences with mixed reality, where you can blend digital objects into the room around you or dial up the immersion in...
- Have more fun with friends in Quest. Whether you’re stepping into an immersive game with people from around the world, watching a live concert...
DeepSeek-R1 achieves performance comparable to OpenAI-o1-1217 on reasoning tasks. To support the research community, we open-source DeepSeek-R1-Zero, DeepSeek-R1, and six dense models (1.5B, 7B, 8B, 14B, 32B, 70B) distilled from DeepSeek-R1 based on Qwen and Llama,” said the developers.
