Skip to main content

Learn

Plain-English guides to the DeepSeek API, R1 reasoning, and the broader OpenAI-compatible LLM ecosystem. No fluff, no affiliate spam.

The DeepSeek API provides cost-effective access to state-of-the-art language models, including DeepSeek-V3 and DeepSeek-R1 reasoning models. It uses OpenAI-compatible endpoints, supports prompt caching for massive token savings, and powers conversational, coding, and mathematical reasoning workflows with low latency and predictable rate limits.

DeepSeek basics

Compare LLMs

Engineering

How models work

Start here

Frequently asked

What models are available on the DeepSeek API?
The DeepSeek API offers DeepSeek-V3 (deepseek-chat) for general and coding tasks, and DeepSeek-R1 (deepseek-reasoner) for step-by-step chain-of-thought mathematical, logic, and code reasoning.
Is the DeepSeek API compatible with OpenAI SDKs?
Yes, the DeepSeek API is fully OpenAI-compatible. You can use standard OpenAI SDKs by setting base_url to https://api.deepseek.com and supplying your DeepSeek API key.
What guides are available in the DeepSeek Learn hub?
The hub covers DeepSeek-R1 reasoning mechanics, API pricing comparisons against GPT-4o, exponential backoff retry patterns, prompt cache optimizations, and complete terminology glossaries.
How does DeepSeek API pricing compare to OpenAI?
DeepSeek is significantly cheaper than GPT-4o, charging approximately $0.14 per million cached input tokens and $0.55 per million uncached inputs on V3, providing up to 90% cost reduction.
How does DeepSeek prompt caching reduce API costs?
DeepSeek automatically caches repeated prompt prefixes. Cache hits cost up to 90% less than cache misses on input tokens, significantly lowering expenses for multi-turn chats and large context windows.
Is the DeepSeek API compatible with the OpenAI SDK?
Yes. DeepSeek endpoints mirror OpenAI syntax. Developers configure the official OpenAI SDK with base_url='https://api.deepseek.com' and their DeepSeek API key.