DataLearner logo
DE

DeepSeek-V4-Pro

Reasoning modelDeepSeek ProDeepSeek V4

DeepSeek V4 Pro

Release date: 2026-04-24Updated: 2026-07-17 21:57:19.252Knowledge cutoff: 2025-0512,448
Live demoGitHubHugging FaceCompare
Parameters
1.6T
Context length
1M
Chinese support
Supported
Reasoning ability

DeepSeek-V4-Pro is a reasoning model from DeepSeek-AI, released on 2026-04-24. It accepts text and image input and returns text output. The cataloged parameter count is 1.6T, with 49B active parameters per inference. The recorded context window is 1M, and the recorded maximum output is 375K. Cataloged capabilities include Reasoning model and Multilingual. The checkpoint is listed under the MIT license, with the license link included in the model references. Use the linked references to confirm current access, licensing, and provider-specific limits.

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

DeepSeek-V4-Pro

Model basics

Reasoning traces
Supported
Thinking modes
Thinking Level · Max (Default)Standard ModeThinking Level · High
Context length
1M tokens
Max output length
375K tokens
Model type
Reasoning model
Modality (in / out)
Text, Image → Text
Release date
2026-04-24
Model file size
No data
MoE architecture
Yes
Total params / Active params
1.6T / 49B
Knowledge cutoff
2025-05
DeepSeek-V4-Pro

Open source & experience

Code license
MIT License
Weights license
MIT License- Commercial use permitted
GitHub repo
GitHub link unavailable
Hugging Face
https://huggingface.co/collections/deepseek-ai/deepseek-v4
Live demo
https://chat.deepseek.com
DeepSeek-V4-Pro

Official resources

Paper
Paper
DataLearnerAI blog
No blog post yet
DeepSeek-V4-Pro

API details

API speed
4/5
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Learn about pricing modes
Standard
TypeConditionInputOutput
Text-$0.435/ 1M$0.870/ 1M
Cache PricingPrompt Cache
TypeTTLWriteRead
Text1d$0.0036/ 1M$0.870/ 1M
DeepSeek-V4-Pro

Benchmark Results

DeepSeek-V4-Pro currently shows benchmark results led by LiveCodeBench (1 / 123, score 93.50), MMLU Pro (11 / 132, score 87.50), SWE-bench Verified (11 / 112, score 80.60). 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
Tool usage

General Knowledge

12 evaluations
Benchmark / mode
Score
Rank/total
GPQA Diamond
Standard Mode
72.90
108 / 187
GPQA Diamond
High
89.10
27 / 187
GPQA Diamond
Max
90.10
23 / 187
MMLU Pro
Standard Mode
82.90
48 / 132
MMLU Pro
High
87.10
13 / 132
MMLU Pro
Max
87.50
11 / 132
LiveBench
Standard Mode
73.58
23 / 115
HLE
Standard Mode
7.70
156 / 172
HLE
High
34.50
77 / 172
HLE
HighTools
44.70
44 / 172
HLE
Max
37.70
67 / 172
HLE
Extra-HighTools
48.20
36 / 172

Coding and Software Engineer

14 evaluations
Benchmark / mode
Score
Rank/total
CodeForces
High
2919
4 / 16
CodeForces
Max
3206
2 / 16
LiveCodeBench
Standard Mode
56.80
76 / 123
LiveCodeBench
High
89.80
6 / 123
LiveCodeBench
Max
93.50
1 / 123
SWE-bench Verified
Standard ModeTools
73.60
45 / 112
SWE-bench Verified
HighTools
79.40
19 / 112
SWE-bench Verified
Extra-HighTools
80.60
11 / 112
SWE-bench Multilingual
Standard ModeTools
69.80
18 / 23
SWE-bench Multilingual
HighTools
74.10
8 / 23
SWE-bench Multilingual
Extra-HighTools
76.20
6 / 23
SWE-Bench Pro - Public
Standard ModeTools
52.10
38 / 54
SWE-Bench Pro - Public
HighTools
54.40
29 / 54
SWE-Bench Pro - Public
Extra-HighTools
55.40
26 / 54

Common Sense Reasoning

1 evaluations
Benchmark / mode
Score
Rank/total
Simple Bench
Standard Mode
50.90
28 / 63

AI Agent - Information Search

2 evaluations
Benchmark / mode
Score
Rank/total
BrowseComp
HighTools
80.40
16 / 53
BrowseComp
Extra-HighTools
83.40
13 / 53

AI Agent - Tool Usage

3 evaluations
Benchmark / mode
Score
Rank/total
Terminal Bench 2.0
Standard ModeTools
59.10
22 / 47
Terminal Bench 2.0
HighTools
63.30
14 / 47
Terminal Bench 2.0
Extra-HighTools
67.90
9 / 47

Math and Reasoning

3 evaluations
Benchmark / mode
Score
Rank/total
IMO-AnswerBench
Standard Mode
35.30
21 / 21
IMO-AnswerBench
High
88
6 / 21
IMO-AnswerBench
Max
89.80
4 / 21

Productivity Knowledge

1 evaluations
Benchmark / mode
Score
Rank/total
GDPval-AA
Extra-HighTools
1554
4 / 21
View benchmark analysisCompare with other models

Compare with other models

DeepSeek-V4-Pro

Publisher

DeepSeek V4 Pro

Model Overview

DeepSeek-V4-Pro is a reasoning model from DeepSeek-AI, released on 2026-04-24.

It accepts text and image input and produces text output. Its cataloged capabilities include Reasoning model and Multilingual. The cataloged parameter count is 1.6T, with 49B active parameters per inference. The recorded context window is 1M, and the recorded maximum output is 375K.

The checkpoint is listed under the MIT license, with the license link included in the model references. The page records 4 API pricing rules from DeepSeek-AI; current provider pricing and conditions should be checked before deployment. The evaluation section contains 36 cataloged benchmark results with their recorded modes and scores. The page links 3 release, model-card, repository, or provider references for checking the underlying claims. Specifications, availability, and prices can change; undisclosed values are intentionally left unstated.

DeepSeek-V4-Pro

FAQ

What is DeepSeek-V4-Pro?

DeepSeek-V4-Pro is a reasoning model from DeepSeek-AI, released on 2026-04-24. It accepts text and image input and returns text output. The cataloged parameter count is 1.6T, with 49B active parameters per inference. The recorded context window is 1M, and the recorded maximum output is 375K. Cataloged capabilities include Reasoning model and Multilingual. The checkpoint is listed under the MIT license, with the license link included in the model references. Use the linked references to confirm current access, licensing, and provider-specific limits.

What input and output modalities does DeepSeek-V4-Pro support?

The current model record lists text and image as input and text as output.

What are the main recorded specifications for DeepSeek-V4-Pro?

The cataloged parameter count is 1.6T, with 49B active parameters per inference. The recorded context window is 1M, and the recorded maximum output is 375K. Fields without a source-backed value remain undisclosed.

Does DeepSeek-V4-Pro have API pricing?

The page records 4 API pricing rules from DeepSeek-AI; current provider pricing and conditions should be checked before deployment.

Are benchmark results available for DeepSeek-V4-Pro?

The evaluation section contains 36 cataloged benchmark results with their recorded modes and scores. Compare only results that use the same benchmark version and evaluation mode.

Is DeepSeek-V4-Pro open source?

The checkpoint is listed under the MIT license, with the license link included in the model references. Review the linked license text before commercial or derivative use.

DataLearner on WeChat

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

DataLearner WeChat QR code