TPSTokens Per Second

DeepSeek R1

DeepSeek🇨🇳 China
active

Open-source reasoning model matching o1 performance at a fraction of the cost.

Compare with other models →
Context window64K tokens
Input / 1M tokens$0.7
Output / 1M tokens$2.5

Version History

DeepSeek-R1major

Full weights open-sourced. Matches OpenAI o1 on math and coding benchmarks at 1/30th the API price.

1.0major

DeepSeek R1 initial release: 671B parameter open-source model with o1-level performance claims, 64K context window, fully visible reasoning tokens, and MIT commercial licensing.

Benchmark Scores

Full leaderboard →
91.4%
AIME 2024
91.0%
AIME 2025
79.5%
GPQA
92.8%
HumanEval
72.6%
LiveCodeBench
97.3%
MATH
90.8%
MMLU
58.0 tokens_per_sec
Speed (tok/s)
57.0%
SWE-bench Verified

Coverage