DeepSeek
First-party API for DeepSeek models
Last reviewed Sep 28, 2026
DeepSeek develops the DeepSeek family of open-weight models and runs a first-party API for them, with OpenAI- and Anthropic-format endpoints, context caching, and time-of-day pricing that halves rates outside weekday peak hours.
LLM API Pricing
Pay-per-token pricing. Prices shown per 1M tokens.
Prices last updated: October 6, 2026
| Model | Input/1M | Output/1M |
|---|---|---|
| $0.300 | $1.20 | |
| $1.32 | $3.96 |
DeepSeek pricing by model
Input, output, and batch rates, plus alternatives, for one model at a time.
Pros & Cons
Advantages
- First-party rates for DeepSeek's own models
- Cache-hit input billed at a fraction of the cache-miss rate
- Off-peak hours cut every rate in half
- Compatible with OpenAI and Anthropic client libraries
Limitations
- Only DeepSeek's own models are offered
- Rates depend on time of day, which complicates cost forecasting
- Per-account concurrency limits apply
- Operated from China; check data-residency requirements before sending sensitive data
Key Features
OpenAI and Anthropic formats
The same models are served on an OpenAI-compatible base URL and an Anthropic-compatible one
Thinking mode
Models support thinking and non-thinking modes selected per request
Context caching
Repeated prompt prefixes are billed at a cache-hit rate well below the cache-miss input rate
Off-peak pricing
Rates are halved outside the published weekday peak hours and on weekends
JSON output and tool calls
Structured JSON output, tool calling, chat prefix completion and FIM completion (beta)
Pricing Options
| Option | Details |
|---|---|
| Per token, peak | Cache-hit input, cache-miss input and output billed per 1M tokens during weekday peak hours |
| Per token, off-peak | Half the peak rates at all other times |
Availability & Support
Support
Documentation, API status page, and FAQ
Company
Published by DeepSeek to identify the company behind this listing.
- Headquarters
- 🇨🇳 Hangzhou, China
Getting Started
- 1
Create an account
Sign up on the DeepSeek Platform and top up a balance
- 2
Create an API key
Generate a key in the platform's API keys page
- 3
Call the API
Point an OpenAI- or Anthropic-compatible client at api.deepseek.com
Frequently Asked Questions
Which models does DeepSeek serve?
Check the pricing table above for the models DeepSeek currently serves and their per-token rates.
How do I get started with DeepSeek?
Create an account, Create an API key, Call the API
What are DeepSeek's main advantages?
DeepSeek's main advantages include: First-party rates for DeepSeek's own models, Cache-hit input billed at a fraction of the cache-miss rate, Off-peak hours cut every rate in half, Compatible with OpenAI and Anthropic client libraries.
What are DeepSeek's limitations?
DeepSeek's main limitations include: Only DeepSeek's own models are offered, Rates depend on time of day, which complicates cost forecasting, Per-account concurrency limits apply, Operated from China; check data-residency requirements before sending sensitive data.
Compare Providers
Find the best prices for the same GPUs and models from other providers