DeepSeek R1

published

DeepSeek's reasoning model

Open-source reasoning model. 90.8% MMLU, 97.3% MATH-500. Chain-of-thought reasoning with MIT license.

Provider deepseek Type reasoning Access open_source Params 671B MoE Context 128k License mit

Benchmarks (2)

Benchmark	Score	Reported By	Source
MMLU	90.8%	vendor	source ↗
MATH-500	97.3%	vendor	source ↗

First open-source model to match o1-level reasoning. The reasoning traces are fully visible and inspectable.

Very high token generation for reasoning (thinking tokens). Slower than non-reasoning models.

Provider deepseek

Released 2025-01-20

Training cutoff 2024-12

Created March 22, 2026

Last reconciled Never