DataLearner logoDataLearnerAI
Latest AI Insights
Model Leaderboards
Benchmarks
Model Directory
Model Comparison
Resource Center
Tools
LanguageEnglish
DataLearner logoDataLearner AI

A knowledge platform focused on LLM benchmarking, datasets, and practical instruction with continuously updated capability maps.

Products

  • Leaderboards
  • Model comparison
  • Datasets

Resources

  • Tutorials
  • Editorial
  • Tool directory

Company

  • About
  • Privacy policy
  • Data methodology
  • Contact

© 2026 DataLearner AI. DataLearner curates industry data and case studies so researchers, enterprises, and developers can rely on trustworthy intelligence.

Privacy policyTerms of service
Page navigation
目录
Model catalogDeepSeek V3.2 Speciale
DE

DeepSeek V3.2 Speciale

推理大模型

DeepSeek V3.2 Speciale

Release date: 2025-12-01更新于: 2026-04-08 15:39:58.060656
Live demoGitHubHugging FaceCompare
Parameters
Not disclosed
Context length
128K
Chinese support
Supported
Reasoning ability

DeepSeek V3.2-Speciale 是 V3.2 的高算力推理变体,放宽输出长度限制以提升复杂推理能力,在 IMO、IOI、ICPC 等顶级竞赛评测中达到金牌水平,性能超过 GPT-5,与 Gemini-3.0-Pro 相当,不支持工具调用,MIT 协议开源。

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 Speciale

Model basics

Reasoning traces
Supported
Thinking modes
Thinking Level · On (Default)Thinking Level · Off
Context length
128K tokens
Max output length
32768 tokens
Model type
推理大模型
Release date
2025-12-01
Model file size
No data
MoE architecture
No
Total params / Active params
0.0B / N/A
Knowledge cutoff
No data
DeepSeek V3.2 Speciale

Open source & experience

Code license
DEEPSEEK LICENSE AGREEMENT
Weights license
DEEPSEEK LICENSE AGREEMENT- 免费商用授权
GitHub repo
GitHub link unavailable
Hugging Face
https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale
Live demo
No live demo
DeepSeek V3.2 Speciale

Official resources

Paper
DeepSeek-V3.2-Speciale 简介与使用说明
DataLearnerAI blog
No blog post yet
DeepSeek V3.2 Speciale

API details

API speed
3/5
No public API pricing yet.
DeepSeek V3.2 Speciale

Benchmark Results

DeepSeek V3.2 Speciale currently shows benchmark results led by AIME2025 (22 / 107, score 96), CodeForces (3 / 10, score 2701), HLE (51 / 131, score 30.60). 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
All modesThinking
Thinking mode details (1)
All thinking modesDefault (On)

综合评估

1 evaluations
Benchmark / mode
Score
Rank/total
HLE
On
30.60
51 / 131

数学推理

1 evaluations
Benchmark / mode
Score
Rank/total
AIME2025
On
96
22 / 107

编程与软件工程

1 evaluations
Benchmark / mode
Score
Rank/total
CodeForces
On
2701
3 / 10
View benchmark analysisCompare with other models
DeepSeek V3.2 Speciale

Publisher

DeepSeek-AI
DeepSeek-AI
View publisher details
DeepSeek V3.2 Speciale

Model Overview

DeepSeek V3.2 系列共包含三个版本:V3.2-Exp、V3.2 正式版和 V3.2-Speciale。V3.2-Exp 是2025年9月发布的实验版本,用于验证新架构;V3.2 正式版于2026年1月发布,是完成完整后训练流程的通用旗舰版本;V3.2-Speciale 是同期发布的高算力推理变体,专为极限推理场景设计,不支持工具调用。

DeepSeek V3.2-Speciale 与 V3.2 正式版共享相同的模型结构,区别在于推理时放宽了输出长度限制,允许模型使用更多的推理 token 来处理复杂问题。这一设计的代价是放弃了工具调用能力,Speciale 只支持纯思考模式,不支持在推理过程中调用外部工具。

在多项顶级学科竞赛评测中,V3.2-Speciale 达到了金牌水平,包括2025年国际数学奥林匹克(IMO)、国际信息学奥林匹克(IOI)、ICPC 世界总决赛以及中国数学奥林匹克(CMO)。在与闭源模型的横向对比中,V3.2-Speciale 在多个推理 benchmark 上超过了 GPT-5,与 Gemini-3.0-Pro 的表现相当。

需要说明的是,这些成绩依赖于更高的推理算力投入和更长的输出长度,实际使用成本高于 V3.2 正式版,也不适合日常对话、编码助手或需要工具调用的 agent 场景。Speciale 的定位是研究和竞赛导向,适合需要在数学推理、代码竞赛等高难度任务上压榨模型能力上限的场景。模型同样以 MIT 协议开源。

DataLearner 官方微信

欢迎关注 DataLearner 官方微信,获得最新 AI 技术推送

DataLearner 官方微信二维码