# DeepSeek > AI research company offering a chat assistant, a desktop harness and a token-billed API for its language models. DeepSeek is an AI research company that builds large language models and offers them through a web chat, a mobile app, a desktop harness and a developer API. DeepSeek is used either by chatting directly, with optional DeepThink and Search modes, or by calling the API, which is billed per million tokens. **Best for:** Developers who want a token-billed LLM API that accepts OpenAI-style and Anthropic-style requests, and people who want a general chat assistant. - Official site: https://www.deepseek.com - Listed in: [Large Language Models & Prompts](https://trendhub.net/ai-tools/category/large-language-models), [AI Chatbots](https://trendhub.net/ai-tools/category/ai-chatbots), [Developer API](https://trendhub.net/ai-tools/category/developer-api) - Last checked: 2026-10-09 (facts taken from deepseek.com) - Source page: https://trendhub.net/ai-tools/deepseek-llm-chat-assistant-developer-api ## Key facts - Platforms: Web, API - Pricing model: Credits - API: Yes - Made by: DeepSeek ## What is DeepSeek? DeepSeek is an AI research company that develops large language models and publishes some of them openly. Its site leads to four products: a web chat with DeepThink and Search switches, a mobile app reached by QR code, a desktop program called Harness, and an API platform for developers. The API documentation lists two models, deepseek-flash and deepseek-v4-pro, each with a context length of one million tokens and a thinking mode that is on by default. Requests can be sent in either OpenAI or Anthropic format through separate base URLs, and supported features include JSON output and tool calls. Usage is charged by the number of input and output tokens, taken from a prepaid balance, with lower rates outside peak hours. The documentation also covers connecting coding agents such as Claude Code and Codex. ## Key features - **Web Chat**: Answers questions in a browser, with switches for DeepThink reasoning and web search next to the prompt box. - **Two API Models**: Exposes deepseek-flash and deepseek-v4-pro, each accepting up to one million tokens of context. - **Thinking Mode**: Runs both models with reasoning switched on by default and lets developers turn it off per request. - **Two Request Formats**: Accepts calls written for the OpenAI format and, through a second base URL, for the Anthropic format. - **Structured Output and Tools**: Returns JSON on request and supports tool calls, plus beta options for prefix and fill-in-the-middle completion. - **Agent Integrations**: Documents how to connect coding agents including Claude Code, Codex, OpenCode and its own Harness. ## Use cases - **Adding an LLM to an Application**: Call the API from existing OpenAI-style client code by changing the base URL and model name. - **Long-Document Processing**: Send large inputs that fit inside the one-million-token context window. - **Running a Coding Agent**: Point a supported agent tool at the API following the integration guides. - **Scheduling Batch Jobs Off-Peak**: Move heavy workloads outside peak hours, when token rates are half the peak rate. - **Everyday Question Answering**: Use the chat in a browser or the mobile app for general queries and web-backed answers. ## Things to know - **Peak and off-peak rates**: Off-peak API prices are half the peak ones. On weekdays the peak windows run from 01:00 to 04:00 UTC and again from 06:00 to 10:00 UTC; weekends and Chinese public holidays count as off-peak. - **Prepaid balance**: Fees are deducted from a topped-up or granted balance, and any granted balance is used first. - **Prices can change**: The documentation says rates may be adjusted and advises checking the pricing page regularly. - **Legacy model names redirect**: Older names such as deepseek-v4-flash are still accepted but are served by DeepSeek-V4.1-Flash and billed at the Flash rate. ## FAQs ### What is DeepSeek? DeepSeek is an AI research company that builds large language models. It offers them as a chat assistant, a desktop harness and a developer API. ### What is DeepSeek used for? DeepSeek is used for conversational question answering in its chat products and for building applications and agents on its API. ### Does DeepSeek have an API? Yes. DeepSeek runs an API platform with public documentation, and the endpoint accepts both OpenAI-format and Anthropic-format requests. ### Which models does the DeepSeek API offer? The DeepSeek API documentation lists deepseek-flash, which runs DeepSeek-V4.1-Flash, and deepseek-v4-pro. Both have a one-million-token context length. ### How is the DeepSeek API billed? DeepSeek charges for tokens, counting input and output together, at rates quoted per one million tokens; the cost comes out of a topped-up or granted balance. See the official pricing page for current rates. ### Does DeepSeek have a mobile app? Yes. The DeepSeek home page offers an app that is downloaded by scanning a QR code, alongside the web chat. ### What is DeepSeek Harness? DeepSeek Harness is a desktop product linked from the home page, and it also appears in the API documentation among the agent integrations. --- Source: https://trendhub.net/ai-tools/deepseek-llm-chat-assistant-developer-api (trendhub)