DeepSeek R1
Open SourceOpen SourceAPI Available
Benchmark Scores
92.5/100MMLU90.8
HumanEval92
Math94.3
Reasoning93.8
Coding91.5
About DeepSeek R1
DeepSeek R1 is a reasoning model that matches OpenAI o1 on math and science benchmarks while being open source and dramatically cheaper. It uses reinforcement learning to develop chain-of-thought reasoning capabilities.
Strengths
- +Matches o1 on math/reasoning
- +Open source chain-of-thought
- +Extremely low cost
- +Strong competition math
Weaknesses
- -High latency with thinking
- -Chinese company concerns
- -Inconsistent on some tasks
Pricing
Per 1M tokens
Input$0.55
Output$2.19
Quick Facts
- Context Window
- 128K
- Parameters
- 671B (37B active)
- Release Date
- 2025-01-20
- Category
- Open Source
Compare Models
See how DeepSeek R1 stacks up against other models