DeepSeek
Picture: DeepSeek

Assistant

DeepSeek: free chat, API prices and models (2026)

DeepSeek is a chat assistant from the Chinese company of the same name. The chat costs nothing; the API is paid by the token and is among the cheapest. Prices below were read on 16 November 2026.

What DeepSeek is

DeepSeek's series of models began in November 2023. Its reasoning model R1, released on 20 January 2025 with open weights, made the DeepSeek app the most downloaded free app on the US App Store within a week. In November 2026 the API offers two models, DeepSeek V4.1 Flash and DeepSeek V4 Pro, both with a context of one million tokens and a thinking mode that is on by default.

Prices

The chat on the website and in the apps is free. The API is billed per million tokens, cheaper off-peak and much cheaper when the start of a request repeats an earlier one and hits the cache. V4.1 Flash costs US$0.15 to 0.30 per million input tokens and US$0.60 to 1.20 per million output tokens; V4 Pro costs US$0.66 to 1.32 and US$1.98 to 3.96. A cache hit costs from US$0.003 per million tokens.

PlanPriceWhat it gives
ChatFreeThe web version and the apps
API: deepseek-flashUS$0.15 to 0.30 per million input tokens, US$0.60 to 1.20 per million output tokensDeepSeek V4.1 Flash, 1 million tokens of context; the lower price applies off-peak
API: deepseek-v4-proUS$0.66 to 1.32 per million input tokens, US$1.98 to 3.96 per million output tokensDeepSeek V4 Pro, without image input

Prices as listed on November 16, 2026

How to start

  1. Open the DeepSeek website or install the app for iOS or Android.
  2. Sign up and start a chat: there is nothing to pay.
  3. For the API, create a key on the DeepSeek Platform and send requests to api.deepseek.com in the OpenAI format, or to api.deepseek.com/anthropic in the Anthropic format.

What people use it for

  • A free chat for everyday questions and writing
  • Long documents, with a context of one million tokens
  • A cheap API for apps and agents; the documentation lists coding agents such as Claude Code and Codex among its integrations
  • Running the models DeepSeek released with open weights, R1 among them, on one's own hardware

Limits and caveats

The API limits parallel requests: 2,500 for V4.1 Flash and 500 for V4 Pro. V4 Pro does not accept images. The prices change between peak and off-peak hours, so a batch job is cheaper when it runs off-peak.

Alternatives

ChatGPT, Gemini and Claude also have free plans, with paid ones above them. Their pages are linked below.

Questions and answers

Is DeepSeek free?

The chat is free on the website and in the apps. The API is paid per token.

How much does the DeepSeek API cost?

From US$0.15 per million input tokens for V4.1 Flash off-peak, and from US$0.003 when the request hits the cache. Output costs from US$0.60 per million tokens.

How long a context does DeepSeek have?

One million tokens for both API models, with up to 384,000 tokens of output.

Similar tools

All tools →