DataLearner logo
DE

DeepSeek V3.2

Reasoning modelDeepSeek VDeepSeek V3.2
Release date: 2025-12-01Updated: 2026-07-17 21:57:32.5906,069
Live demoGitHubHugging FaceCompare
Parameters
671B
Context length
128K
Chinese support
Supported
Reasoning ability

DeepSeek V3.2 is a reasoning model from DeepSeek-AI, released on 2025-12-01. It accepts text input and returns text output. The cataloged parameter count is 671B, with 37B active parameters per inference. The recorded context window is 128K, and the recorded maximum output is 8K. Cataloged capabilities include Reasoning model and Multilingual. The checkpoint is listed under the Deepseek License Agreement 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 V3.2

Model basics

Reasoning traces
Supported
Thinking modes
Thinking Mode (Default)Standard Mode
Context length
128K tokens
Max output length
8K tokens
Model type
Reasoning model
Modality (in / out)
Text → Text
Release date
2025-12-01
Model file size
1.34TB
MoE architecture
Yes
Total params / Active params
671B / 37B
Knowledge cutoff
No data
DeepSeek V3.2

Open source & experience

DeepSeek V3.2

Official resources

Paper
Paper
DataLearnerAI blog
DataLearnerAI blog
DeepSeek V3.2

API details

API speed
3/5
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Learn about pricing modes
Standard
TypeConditionInputOutput
Text-$0.280/ 1M$0.420/ 1M
Cache PricingPrompt Cache
TypeTTLWriteRead
Text-$0.280/ 1M$0.028/ 1M
DeepSeek V3.2

Benchmark Results

DeepSeek V3.2 currently shows benchmark results led by LiveCodeBench (21 / 123, score 83.30), AIME2025 (30 / 107, score 93.10), τ²-Bench (14 / 43, score 80.30). 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
Tool usage

General Knowledge

6 evaluations
Benchmark / mode
Score
Rank/total
GPQA Diamond
Thinking Mode
82.40
69 / 187
LiveBench
Standard Mode
51.84
87 / 115
LiveBench
Thinking Mode
62.20
58 / 115
ARC-AGI
Thinking Mode
57
41 / 68
HLE
Thinking Mode
25.10
102 / 172
ARC-AGI-2
Thinking Mode
4
51 / 62

Coding and Software Engineer

5 evaluations
Benchmark / mode
Score
Rank/total
CodeForces
Thinking Mode
2386
11 / 16
LiveCodeBench
Thinking Mode
83.30
21 / 123
SWE-bench Verified
73.10
49 / 112
SWE-bench Verified
Thinking Mode
70.20
60 / 112
SWE-Bench Pro - Public
Thinking Mode
40.90
49 / 54

Math and Reasoning

3 evaluations
Benchmark / mode
Score
Rank/total
AIME2025
Thinking Mode
93.10
30 / 107
AIME 2026
Thinking Mode
92.70
9 / 18
FrontierMath - Tier 4
Thinking Mode
2.10
56 / 80

Agent Level Benchmark

1 evaluations
Benchmark / mode
Score
Rank/total
τ²-Bench
80.30
14 / 43

AI Agent - Information Search

1 evaluations
Benchmark / mode
Score
Rank/total
BrowseComp
Thinking Mode
51.40
42 / 53

AI Agent - Tool Usage

1 evaluations
Benchmark / mode
Score
Rank/total
Terminal Bench 2.0
46.40
40 / 47

Claw-style Agent Evaluation

2 evaluations
Benchmark / mode
Score
Rank/total
Pinch Bench
Thinking ModeTools
84.30
18 / 37
Claw Bench
Thinking ModeTools
79
21 / 29
View benchmark analysisCompare with other models

Compare with other models

DeepSeek V3.2

Publisher

DeepSeek V3.2

Model Overview

DeepSeek V3.2 is a reasoning model from DeepSeek-AI, released on 2025-12-01.

It accepts text input and produces text output. Its cataloged capabilities include Reasoning model and Multilingual. The cataloged parameter count is 671B, with 37B active parameters per inference. The recorded context window is 128K, and the recorded maximum output is 8K.

The checkpoint is listed under the Deepseek License Agreement license, with the license link included in the model references. The page records 6 API pricing rules from DeepSeek-AI; current provider pricing and conditions should be checked before deployment. The evaluation section contains 19 cataloged benchmark results with their recorded modes and scores. The page links 5 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 V3.2

FAQ

What is DeepSeek V3.2?

DeepSeek V3.2 is a reasoning model from DeepSeek-AI, released on 2025-12-01. It accepts text input and returns text output. The cataloged parameter count is 671B, with 37B active parameters per inference. The recorded context window is 128K, and the recorded maximum output is 8K. Cataloged capabilities include Reasoning model and Multilingual. The checkpoint is listed under the Deepseek License Agreement 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 V3.2 support?

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

What are the main recorded specifications for DeepSeek V3.2?

The cataloged parameter count is 671B, with 37B active parameters per inference. The recorded context window is 128K, and the recorded maximum output is 8K. Fields without a source-backed value remain undisclosed.

Does DeepSeek V3.2 have API pricing?

The page records 6 API pricing rules from DeepSeek-AI; current provider pricing and conditions should be checked before deployment.

Are benchmark results available for DeepSeek V3.2?

The evaluation section contains 19 cataloged benchmark results with their recorded modes and scores. Compare only results that use the same benchmark version and evaluation mode.

Is DeepSeek V3.2 open source?

The checkpoint is listed under the Deepseek License Agreement 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