AI tools explained · Open models
What is DeepSeek? The AI lab and its models explained
What is DeepSeek? DeepSeek is an AI lab whose models, from R1 to V4.1, are available through a chat app, an API and open-source projects on GitHub.
| Made by | DeepSeek |
|---|---|
| Type | Open models |
| Pricing | Free web chat and app; API billed by usage (see DeepSeek API pricing) |
| Official site | deepseek.com |
DeepSeek is the AI lab many developers look to when they want strong models they can run or inspect themselves. It also matters for agents: open-source tools like Hermes Agent let you plug in whichever model provider you prefer. Here is what DeepSeek offers.
What is DeepSeek?
DeepSeek is an AI lab that builds large language models and publishes its releases. Its website lists a line of models from DeepSeek V3 and R1 through V3.1, V3.2, V4 and V4.1, with a banner noting that DeepSeek V4.1 Flash was updated to version V4.1-Flash-1017 (DeepSeek, checked 2 November 2026).
Where can you use DeepSeek?
| Product | What it is |
|---|---|
| DeepSeek Web | Chat in the browser at chat.deepseek.com |
| DeepSeek App | Mobile app |
| API platform | Developer API; now supports web search and fetch |
| DeepSeek Harness | Agent tool and desktop app (“Everything is a Plugin”) |
| Source | DeepSeek |
What does DeepSeek publish on GitHub?
DeepSeek’s GitHub organisation had about 113,000 followers when we checked (GitHub). Its pinned projects include DeepSeek Harness, MIT-licensed with about 269,000 stars, and DeepInfer, an Apache-2.0 inference framework for DeepSeek models with about 25,000 stars. It also publishes lower-level performance libraries such as FlashMLA, DeepEP and DeepGEMM under the MIT license.
How much does DeepSeek cost?
The web chat and app are free to use, and the API is billed by usage; DeepSeek links to a separate API pricing page and says the V4.1 Flash update lets tasks finish “at lower cost and faster speed” (DeepSeek). Check the API pricing page for current rates.
How does DeepSeek compare with Claude and ChatGPT?
Claude and ChatGPT are polished assistants from companies that keep their main models closed. DeepSeek’s appeal for builders is openness: much of its tooling is published under permissive licenses, and its API is a common choice inside agent frameworks. The trade-off is that you do more of the setup yourself.
Frequently asked questions
What is the latest DeepSeek model?
DeepSeek's site lists DeepSeek V4.1 as its newest release, with V4.1 Flash updated to version V4.1-Flash-1017 (checked November 2026). Earlier releases include V4, V3.2, V3.1, R1 and V3.
Is DeepSeek open source?
DeepSeek publishes many projects on GitHub under MIT and Apache-2.0 licenses, including DeepSeek Harness and DeepInfer. Check each model's own license before using its weights.
What is DeepSeek Harness?
DeepSeek Harness is DeepSeek's agent tool, described on GitHub as 'Everything is a Plugin.' It is MIT-licensed and also offered as a desktop app.
Does DeepSeek have an API?
Yes. DeepSeek runs an API platform, and its site says the API now supports web search and fetch.
Sources
- DeepSeek – DeepSeek, 2026-11
- deepseek-ai on GitHub – GitHub, 2026-11
More AI tools explained
- What is ChatGPT? OpenAI's AI assistant and its plans explained
- What are ChatGPT dots? OpenAI's always-on agents explained
- What is Claude? Anthropic's AI assistant explained
- What is Claude Code? Anthropic's coding agent explained
- What is Codex? OpenAI's coding agent explained
- What is Cursor AI? The AI code editor explained
- What is Google Gemini? Google's AI assistant explained
- What is Grok AI? xAI's assistant with live X search explained
- What is Hermes Agent? Nous Research's open-source agent
- What is Microsoft Copilot? AI in Word, Excel and Outlook
- What is Perplexity AI? The AI answer engine explained