DeepSeek API on Novita AI
Reason, code, create, and understand documents. Find the right DeepSeek model for your task. Start building with one API.
Built for your next task
Start with your task. Find your model.
The DeepSeek collection
All models · 9 versions · Showing 6
Everyday choicedeepseek/deepseek-v4.1-flash
Text and image input with chat, reasoning, and tool calling. A versatile starting point for everyday tasks.
1017 versiondeepseek/deepseek-v4.1-flash-1017
The 1017 version of V4.1 Flash, with text and image input for chat and multimodal understanding.
0813 versiondeepseek/deepseek-v4-pro-0813
The 0813 version of V4 Pro, with reasoning, structured output, and tool calling. Compare it with other Pro versions.
Coding & agentsdeepseek/deepseek-v4-flash-0731
The 0731 revision of V4 Flash, suited to coding, reasoning, and agent workflows with tool calling.
Complex tasksdeepseek/deepseek-v4-pro
Built for complex coding, mathematical reasoning, and multi-step agents, with long context and structured output.
Document specialistdeepseek/deepseek-ocr-2
Extract text and reading order from images and complex documents for search, organization, and knowledge bases.
On-demand deploymentdeepseek/deepseek-r1-0528-qwen3-8b
R1 reasoning distilled into Qwen3 8B. Runs on on-demand deployment instead of the Serverless API.
Reasoningdeepseek/deepseek-r1-0528
An R1 reasoning model for math, coding, and logic tasks that require multi-step analysis.
R1 familydeepseek/deepseek-r1-turbo
Another R1 serving option with reasoning and tool calling. Choose based on your task and context needs.
USD / 1M tokens · Live pricing from the Novita model catalog.
Not sure where to start? Try Flash for everyday tasks, R1 or Pro for complex reasoning, and OCR 2 for scanned documents.
Less setup. More building.
Use the OpenAI SDK you already know. Connect to the DeepSeek model you need.
Open the console and create your API key.
Copy the model ID and set your API endpoint.
Check your balance and model pricing, then send a request.
# pip install openai
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["NOVITA_API_KEY"],
base_url="https://api.novita.ai/openai",
)
response = client.chat.completions.create(
model="deepseek/deepseek-v4.1-flash",
messages=[{"role": "user", "content": "Explain how an API works."}],
max_tokens=2048,
)
print(response.choices[0].message.content)OpenAI-compatible API · Set your key via an environment variable
Expand your toolkit
Add the right capabilities to your stack.
Explore an alternative for complex coding and multi-step agent workflows.
256K contextView model
Image & video understandingGo beyond text with a model that understands images and video.
128K contextView modelKnowledge & semantic searchEmbed your documents, retrieve relevant knowledge, then let DeepSeek generate an answer.
32K contextView modelStart with your task: try V4.1 Flash for everyday chat, compare R1 and V4 Pro for complex reasoning, math, and coding, or choose OCR 2 for document recognition. Then evaluate context length, pricing, and results on your own workload.
This page lists every DeepSeek model currently available in the Novita model catalog, including models that run on on-demand deployment. Retired versions are not listed.
Yes. Set base_url to https://api.novita.ai/openai, use your Novita API key, and specify the model ID from its detail page. Support for tool calling, structured output, and other features varies by model.
Serverless models are billed by token usage, including input and output, with no deployment to manage. Cache-read pricing varies by model. Models marked On-demand deployment run on a dedicated deployment instead; see their model pages for options.
Yes. Use an embedding model to index and retrieve relevant text, then pass the retrieved context to DeepSeek to generate an answer. Scanned documents can go through OCR first. Evaluate answer quality against your own business data.
Your next idea starts here
From your first API call to your next AI application.