Model Library/Deepseek V3.2
deepseek/deepseek-v3.2

Deepseek V3.2

deepseek/deepseek-v3.2
We introduce DeepSeek-V3.2, a next-generation foundation model designed to unify high computational efficiency with state-of-the-art reasoning and agentic performance. DeepSeek-V3.2 is built upon three core technical breakthroughs: • DeepSeek Sparse Attention (DSA): A new highly efficient attention mechanism that significantly reduces computational overhead while preserving model quality, purpose-built for long-context reasoning and high-throughput workloads. • Scalable Reinforcement Learning Framework: DeepSeek-V3.2 leverages a robust RL training protocol and expanded post-training compute to reach GPT-5-level performance. Its high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and demonstrates reasoning capabilities comparable to Gemini-3.0-Pro. • Large-Scale Agentic Task Synthesis Pipeline: To enable reliable tool-use and multi-step decision-making, we develop a novel agentic data synthesis pipeline that generates high-quality interactive reasoning tasks at scale, greatly enhancing the model’s

Features

Serverless API

Docs

deepseek/deepseek-v3.2 is available via Novita's serverless API, where you pay per token. There are several ways to call the API, including OpenAI-compatible endpoints with exceptional reasoning performance.

Available Serverless

Run queries immediately, pay only for usage

Input$0.28 / M Tokens
Output$0.42 / M Tokens

Use the following code examples to integrate with our API:

1from openai import OpenAI
2
3client = OpenAI(
4    api_key="<Your API Key>",
5    base_url="https://api.novita.ai/openai"
6)
7
8response = client.chat.completions.create(
9    model="deepseek/deepseek-v3.2",
10    messages=[
11        {"role": "system", "content": "You are a helpful assistant."},
12        {"role": "user", "content": "Hello, how are you?"}
13    ],
14    max_tokens=65536,
15    temperature=0.7
16)
17
18print(response.choices[0].message.content)

Info

Provider
DeepSeek
Quantization
fp8

Supported Functionality

Context Length
163840
Max Output
65536
Serverless
Supported
Reasoning
Supported
Anthropic API
Supported
Input Capabilities
text
Output Capabilities
text