DeepSeek API

What it does

The DeepSeek API accepts chat requests as a list of messages (system and user roles) sent over HTTPS with an API key, and returns model-generated text. Responses can be streamed or returned in full. Two model names are available: deepseek-flash and deepseek-v4-pro. Optional parameters include a thinking mode and a reasoning-effort setting.

Developers point an existing OpenAI or Anthropic SDK, or any software compatible with those formats, at DeepSeek's base URL and swap in a DeepSeek key and model name. That compatibility keeps code changes small when moving an application over. Agent and coding tools such as Claude Code, GitHub Copilot and OpenCode can also use DeepSeek as their backend model without custom code. A DeepSeek Harness is available in developer preview for agent harness developers.

It suits developers and teams building applications or tooling on a hosted model. An API key is required, and you call the service from code, curl or a compatible tool rather than through an end-user app. For rates, rate limits and context sizes, see the Models & Pricing page. Older flash model names are still accepted but are now served by the current Flash model.

What it can do

• Chat completions endpoint taking system and user messages
• OpenAI-format and Anthropic-format API access via separate base URLs
• Streaming or non-streaming responses via the stream parameter
• Optional thinking mode with a reasoning_effort setting
• Use as the backend model for agent and coding assistant tools
• Two named models: deepseek-flash and deepseek-v4-pro

Works with

• OpenAI SDK
• Anthropic SDK
• Claude Code
• GitHub Copilot
• OpenCode

Example workflows

Switch an existing OpenAI-based app to DeepSeek

Illustrative example. Inputs: an app using the OpenAI SDK and a DeepSeek API key. Steps: set base_url to https://api.deepseek.com, set the key, change the model to deepseek-flash, send a test chat request. Output: a chat completion from the DeepSeek model.

Use DeepSeek as a coding assistant backend

Illustrative example. Inputs: a DeepSeek API key and a supported tool such as Claude Code or OpenCode. Steps: follow the Agent Integrations Guide to configure the tool's model endpoint and key. Output: the tool's responses generated by DeepSeek.

Before you choose

• Legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are retired and now served by the current Flash model, billed at the Flash price.
• DeepSeek Harness is only a developer preview.
• Review the Models & Pricing page for current rates before estimating costs.

Listed under

• Coding
• Coding agents
• Coding agent
• Developer API
• Developers
• Image understanding
• LLM API

Get started

Apply for a key on the platform, then set the base URL and model name in your SDK.

1. Apply for an API key on the DeepSeek platform
2. Set the base URL (OpenAI or Anthropic format) and key in your SDK or tool
3. Send a test request to /chat/completions with model deepseek-flash

The published information link to a Models & Pricing page, but no rates, free allowance or billing basis were confirmed. Legacy Flash names are billed at the Flash price.

We couldn’t find published prices for DeepSeek API. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Your First API Call | DeepSeek API Docs ↗api-docs.deepseek.com
• Your First API Call | DeepSeek API Docs ↗api-docs.deepseek.com

Company website

platform.deepseek.com: product, pricing and policy pages

Web archive

When the site first appeared

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.