DE

DeepSeek-R1

Reasoning modelDeepSeek R1

DeepSeek-R1

Release date: 2025-01-20Updated: 2025-03-21 11:14:181,760
Live demoGitHubHugging FaceCompare
Parameters
671B
Context length
128K
Chinese support
Supported
Reasoning ability

DeepSeek-R1 is an AI model published by DeepSeek-AI, released on 2025-01-20, for Reasoning model, with 671B parameters, and 128K context length, requiring about 134GB storage, under the MIT License license, with a 97.30 score on MATH-500.

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

DeepSeek-R1

Model basics

Reasoning traces
Supported
Thinking modes
Thinking modes not supported
Context length
128K tokens
Max output length
No data
Model type
Reasoning model
Modality (in / out)
No data
Release date
2025-01-20
Model file size
134GB
MoE architecture
No
Total params / Active params
671B / N/A
Knowledge cutoff
No data
DeepSeek-R1

Open source & experience

Code license
Weights license
MIT License- 免费商用授权
GitHub repo
GitHub link unavailable
Live demo
No live demo
DeepSeek-R1

Official resources

DeepSeek-R1

API details

API speed
No data
No public API pricing yet.
DeepSeek-R1

Benchmark Results

DeepSeek-R1 currently shows benchmark results led by MMLU (8 / 65, score 90.80), MMLU Pro (37 / 126, score 84), MATH-500 (13 / 44, score 97.30). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking

General Knowledge

5 evaluations
Benchmark / mode
Score
Rank/total
90.80
8 / 65
84
37 / 126
71.50
104 / 179
69.41
22 / 52
15.80
55 / 65

Common Sense

1 evaluations
Benchmark / mode
Score
Rank/total
30.10
22 / 45

Coding and Software Engineer

2 evaluations
Benchmark / mode
Score
Rank/total
65.90
59 / 120
49.20
92 / 108

Math and Reasoning

3 evaluations
Benchmark / mode
Score
Rank/total
97.30
13 / 44
79.80
28 / 62
70
73 / 106

Writing and Creative Capabilities

1 evaluations
Benchmark / mode
Score
Rank/total
84.60
11 / 23

Compare with other models

No curated comparisons for this model yet.

Want a custom combination? Open the compare tool

DeepSeek-R1

Publisher

DeepSeek-R1

Model Overview

DeepSeek-R1 is an AI model published by DeepSeek-AI, released on 2025-01-20, for Reasoning model, with 671B parameters, and 128K context length, requiring about 134GB storage, under the MIT License license, with a 97.30 score on MATH-500.

DataLearner on WeChat

Follow DataLearner on WeChat for AI model updates and research notes.

DataLearner WeChat QR code