# DeepSeek

> Free AI chat app and low-cost API running DeepSeek's open-weights V4 models

- Category: [AI Assistants](https://finderatlas.com/categories/ai-assistants)
- Pricing: Freemium
- Website: https://www.deepseek.com
- Platforms: Web, iOS, Android, macOS, Windows, API
- API: available
- Mobile app: available
- Made by: DeepSeek, founded 2023, Hangzhou, China
- Last updated: 2026-10-20
- FinderAtlas page: https://finderatlas.com/tools/deepseek

## Overview

DeepSeek is an AI assistant and model provider from the Hangzhou-based company of the same name. Its chat service at chat.deepseek.com and its iOS and Android apps are free to use, and they run the company's current V4 model family. Developers can call the same models through a paid, pay-as-you-go API at platform.deepseek.com.

## Key features

- V4 model family
- Thinking mode
- Open weights
- OpenAI and Anthropic compatible API
- Tool calls and JSON output
- Search and Fetch APIs
- Agent tool integrations
- DeepSeek Harness
- Vision on Flash

## Pricing

- **Chat and App** ($0): Free access to DeepSeek models on web, iOS and Android
- **API: deepseek-flash** ($0.15/1M input tokens): Pay as you go; $0.15 cache-miss input and $0.60 output per 1M tokens off-peak, double at peak
- **API: deepseek-v4-pro** ($0.66/1M input tokens): Pay as you go; $0.66 cache-miss input and $1.98 output per 1M tokens off-peak, double at peak

## Best for

- Low-cost API integration
- Coding and math reasoning
- Self-hosting open-weights models
- Free everyday chat

## Who uses it

- Software developers
- Researchers and analysts
- Students
- AI engineers
- Cost-conscious teams

## Strengths

- Web chat and mobile apps are free
- API output costs from $0.60 per 1M tokens (Flash, off-peak), well below most frontier providers
- Open weights for V4 and earlier models on Hugging Face
- 1M-token context and 384K max output on both API models
- API works with both OpenAI and Anthropic SDKs and several coding agents

## Limitations

- Peak-hour API rates are double the off-peak rates on weekday mornings UTC
- deepseek-v4-pro does not support vision input
- No image generation
- Data is processed by a company based in China, which some organizations cannot accept

## Not recommended for

- Image generation
- Users who cannot send data to servers in China

## Integrations

- OpenAI SDK
- Anthropic SDK
- Claude Code
- Codex
- OpenCode
- Hugging Face

## FinderAtlas editorial scores (out of 10)

- Reasoning & Logic: 8.8/10
- Coding & Syntax: 8.7/10
- Cost Efficiency: 9.6/10
- Long Context: 8.8/10
- Ease of Use: 8/10

## Alternatives

- [Microsoft Copilot](https://finderatlas.com/tools/microsoft-copilot): Microsoft's AI assistant for chat, writing and research across web, apps and Microsoft 365 ([compare](https://finderatlas.com/compare/deepseek-vs-microsoft-copilot))
- [Poe](https://finderatlas.com/tools/poe): Chat with many AI models, image tools and custom bots through one account ([compare](https://finderatlas.com/compare/deepseek-vs-poe))
- [Perplexity](https://finderatlas.com/tools/perplexity): AI answer engine that searches the live web and cites its sources ([compare](https://finderatlas.com/compare/deepseek-vs-perplexity))
- [Msty](https://finderatlas.com/tools/msty): Private AI workspace for chatting with local and online models, with your data under your control ([compare](https://finderatlas.com/compare/deepseek-vs-msty))
- [Google Gemini](https://finderatlas.com/tools/gemini): Google's AI assistant for chat, research, image and video creation, and Workspace apps ([compare](https://finderatlas.com/compare/deepseek-vs-gemini))
- [Mistral Vibe (formerly Le Chat)](https://finderatlas.com/tools/mistral-le-chat): Mistral AI's chat assistant and coding agent for everyday work and software tasks ([compare](https://finderatlas.com/compare/deepseek-vs-mistral-le-chat))
- [ChatGPT](https://finderatlas.com/tools/chatgpt): OpenAI's AI assistant for writing, research, coding, image generation and voice chat ([compare](https://finderatlas.com/compare/chatgpt-vs-deepseek))
- [Claude](https://finderatlas.com/tools/claude): Anthropic's AI assistant for writing, analysis, coding and working with files ([compare](https://finderatlas.com/compare/claude-vs-deepseek))

## Recent updates

- 2026-10-17: Updated Flash model now served by deepseek-flash
- 2026-04-24: Released V4 Pro and Flash with open weights and a 1M-token context

## FAQ

**Are DeepSeek models open source?**

DeepSeek publishes open weights for its models, including V4, on Hugging Face, so they can be downloaded and run on your own hardware. The hosted chat and API are separate services run by DeepSeek.

**Is DeepSeek free to use?**

The web chat at chat.deepseek.com and the iOS and Android apps are free. The API is pay as you go, starting at $0.15 per 1M cache-miss input tokens for deepseek-flash off-peak.

**What is the difference between peak and off-peak API pricing?**

Off-peak rates are half of peak rates. Peak hours are 01:00-04:00 and 06:00-10:00 UTC on weekdays, excluding Chinese public holidays. All other times, including weekends, are off-peak.

**Which models does the DeepSeek API offer?**

Two: deepseek-flash, served by DeepSeek-V4.1-Flash-1017, and deepseek-v4-pro, served by DeepSeek-V4-Pro-0813. Both have a 1M-token context window and up to 384K output tokens. Older names such as deepseek-v4-flash are still accepted and routed to Flash.

**Can I use DeepSeek with the OpenAI or Anthropic SDK?**

Yes. Set the base URL to https://api.deepseek.com for the OpenAI format or https://api.deepseek.com/anthropic for the Anthropic format, and use a DeepSeek API key.

**Does DeepSeek have a mobile app?**

Yes. The official download page offers iOS and Android versions of the DeepSeek app, and a Harness desktop client for macOS and Windows.