DeepSeek API

What is DeepSeek API?

The DeepSeek API gives developers HTTP access to DeepSeek language models, billed by the million input and output tokens. The documentation lists the deepseek-v4-pro model alongside a Flash model, with a 1M-token context window and output of up to 384K tokens. It accepts requests in OpenAI format at api.deepseek.com and in Anthropic format at api.deepseek.com/anthropic, so existing SDK code works after swapping the base URL and key. Models support JSON output, tool calls, the Responses API, and a thinking mode with low, high and max reasoning effort, which suits coding agents and tools like Codex. The docs also describe beta chat prefix completion and fill-in-the-middle completion. Billing separates cache-hit from cache-miss input and halves the price off-peak. The Chinese pricing page lists deepseek-v4-pro at 4.5元 per 1M cache-miss input tokens off-peak and 13.5元 per 1M output tokens, but the changelog says Flash prices were cut in September 2026, so check the live page.

Use cases

• Add an LLM to an app through an API
• Build tool-calling agents
• Return structured JSON output
• Complete code with fill-in-the-middle
• Analyze long documents in a 1M context
• Use DeepSeek as the backend for Codex

Who it's for

• Developers
• AI engineers
• Agent builders
• Startups

Why use DeepSeek API

• The OpenAI and Anthropic-compatible endpoints let you switch an existing client by changing a base URL and key.
• Off-peak prices are half of peak prices, so queued jobs can run at lower cost.
• Cache-hit input is billed far below cache-miss input, which helps repeated prompts that share a prefix.
• Choosing low, high or max reasoning effort lets you match spend to the difficulty of each task.

DeepSeek API features

• OpenAI-format and Anthropic-format base URLs
• 1M-token context and 384K maximum output
• Thinking and non-thinking modes
• Low, high and max reasoning effort
• JSON output and tool calls
• Native OpenAI Responses API support
• Chat prefix completion (beta)
• Fill-in-the-middle completion in non-thinking mode (beta)
• Peak and off-peak pricing with cache-hit discounts
• Concurrency limits of 500 to 2500 depending on model

DeepSeek API integrations

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

How to get started with DeepSeek API

1. Apply for an API key on the DeepSeek platform and top up your balance.
2. Set your OpenAI or Anthropic SDK base URL to the DeepSeek endpoint and pass the key.
3. Choose a model name such as deepseek-v4-pro and enable or disable thinking.
4. Set a reasoning effort of low, high or max for the task.
5. Send requests; costs are deducted from your balance, using granted balance first.

DeepSeek API limitations

• Prices are listed in 元 on the Chinese pricing page and can change without notice
• Fill-in-the-middle only works in non-thinking mode
• Legacy model names such as deepseek-v4-flash are routed to the newer Flash model
• Concurrency is capped at 500 for V4-Pro

DeepSeek API pricing

DeepSeek API has no free plan.

DeepSeek API plans

• deepseek-v4-pro

  4.5元-9.0元 per 1M input tokens

  cache miss, off-peak to peak; usage-based

  • Cache-hit input 0.15元 off-peak, 0.30元 peak
  • Output 13.5元 off-peak, 27.0元 peak
  • Concurrency limit 500

• deepseek-flash

  Price on vendor site

  usage-based

  • 1M context and 384K max output
  • Concurrency limit 2500
  • Prices cut with V4.1 Flash release

Plans as listed on the official pricing page on 24 Oct 2026. Prices can change and may vary by country or exclude tax. Check the official pricing page.

DeepSeek API alternatives

The main DeepSeek API alternatives listed on Debuly are OpenAI API, Mistral AI, Together AI, Groq, DeepSeek and OmniRoute. The table compares pricing and platforms; order is A–Z.

1. 

   DeepSeek

   AI lab and chat assistant with a pay-per-token API for its V4 Flash and V4 Pro models, plus an app and web chat.

   Chat assistantsDeveloper platformsCoding
   FreemiumVisit DeepSeek (official site)
2. 

   Groq

   GroqCloud inference API that serves open models such as GPT-OSS and Whisper on Groq's LPU hardware, billed per token.

   Developer platformsAudio and voiceCoding
   FreemiumVisit Groq (official site)
3. 

   Mistral AI

   AI company offering open-weight models, the Vibe assistant, Studio and an API, with a free plan and paid tiers.

   Developer platformsChat assistantsCoding
   FreemiumVisit Mistral AI (official site)
4. 

   OmniRoute

   Free open-source AI gateway that puts 352 providers behind one OpenAI-compatible endpoint with automatic fallback.

   Developer platformsCodingAutomation and agents
   Open sourceVisit OmniRoute (official site)
5. 

   OpenAI API

   Developer API for OpenAI's GPT models, image generation, realtime voice and transcription, priced per million tokens with batch and flex options.

   Developer platformsCodingAudio and voice
   PaidVisit OpenAI API (official site)
6. 

   Together AI

   AI cloud for running open-source models through serverless and batch inference, dedicated endpoints, GPU clusters and fine-tuning.

   Developer platformsCoding
   PaidVisit Together AI (official site)

DeepSeek API FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: from api-docs.deepseek.com · fetched 23 Oct 2026 · Vendor trademark, used only to identify the product; removed on request
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.