Freemium. Chatting in the official DeepSeek app is free. The API is billed per token in US dollars per 1M tokens, deducted from a topped-up or granted balance. deepseek-flash (DeepSeek-V4.1-Flash) costs $0.15–$0.30 per 1M input tokens on a cache miss and $0.60–$1.20 per 1M output tokens, depending on off-peak or peak hours; deepseek-v4-pro is listed at $0.66–$1.32 input and $1.98–$3.96 output, but its requests have been routed to V4.1-Flash at Flash rates since 14 September 2026.Official source 1 ↗ · Official source 2 ↗ · Official source 3 ↗
Available as
The DeepSeek assistant is available as DeepSeek Web and as mobile apps for iOS and Android, with chat history synced across platforms. · DeepSeek Harness, DeepSeek's agent app, has a desktop version for Apple silicon Macs running macOS 13 or later and for Windows. · The API is reached at https://api.deepseek.com in OpenAI format or https://api.deepseek.com/anthropic in Anthropic format, so OpenAI and Anthropic SDKs and compatible software can use it.Official source 1 ↗ · Official source 2 ↗ · Official source 3 ↗Official source 1 ↗Official source 1 ↗ · Official source 2 ↗
Primary group
Research & Data
What it does
DeepSeek is an AI research company that develops large language models and releases them as open source. Its official assistant lets people chat with DeepSeek models on the web and in iOS and Android apps, with web search, a DeepThink reasoning mode and file upload with text extraction. Developers can call the same models through a pay-as-you-go API that is compatible with the OpenAI and Anthropic formats. The current API model, DeepSeek-V4.1-Flash, accepts images as well as text.
Chat with web search and DeepThink. The DeepSeek assistant answers questions in a chat, with a web search mode and a DeepThink reasoning mode. It runs on DeepSeek's latest flagship model.Official source 1 ↗ · Official source 2 ↗ · Official source 3 ↗
File upload and text extraction. You can upload files in the app and have their text extracted for the conversation.Official source 1 ↗
Synced chat history and account login. Sign in with email, a Google account or an Apple ID, and chat history syncs across platforms. Chat history can be managed, copied or deleted in settings.Official source 1 ↗ · Official source 2 ↗
OpenAI- and Anthropic-compatible API. The DeepSeek API accepts the OpenAI and Anthropic request formats, so existing SDKs work by changing the base URL and API key. Both API models support JSON output, tool calls, the Responses API and the Anthropic API, with a 1M-token context and up to 384K output tokens.Official source 1 ↗ · Official source 2 ↗
Thinking and non-thinking modes. The API models support a non-thinking mode and a thinking mode, which is the default.Official source 1 ↗
Image understanding in the API. deepseek-flash accepts images alongside text, so it can describe pictures, read text from screenshots and analyse charts. It takes JPEG, PNG, GIF and WebP; deepseek-v4-pro does not support vision.Official source 1 ↗ · Official source 2 ↗
Backend for coding agents. Coding and agent tools such as Claude Code, GitHub Copilot and OpenCode can use DeepSeek as their backend model without code. DeepSeek Harness is in developer preview for agent harness developers.Official source 1 ↗
DeepSeek Harness for desktop. DeepSeek Harness on desktop works in the background, edits local files and handles long tasks.Official source 1 ↗
Open-source models. DeepSeek releases its models as open source, including DeepSeek-V4, DeepSeek-R1 and DeepSeek-Coder; DeepSeek-V4.1-Flash is published on Hugging Face with a technical report.Official source 1 ↗ · Official source 2 ↗
Pricing & free limits
DeepSeek app (web, iOS, Android)
$0Free
The official assistant for chatting with DeepSeek models, with web search, DeepThink and file upload. The January 2025 app announcement describes it as free with no ads and no in-app purchases; no paid consumer tier is listed on the checked pages.Official source 1 ↗ · Official source 2 ↗
API: deepseek-flash (DeepSeek-V4.1-Flash)
$0.15–$0.30 per 1M input tokens (cache miss); $0.60–$1.20 per 1M output tokensPay as you go, deducted from a topped-up or granted balance
Cache-hit input costs $0.003 off-peak or $0.006 at peak per 1M tokens; cache-miss input $0.15 off-peak or $0.30 at peak; output $0.60 off-peak or $1.20 at peak. 1M-token context, up to 384K output tokens, image input and a concurrency limit of 2500.Official source 1 ↗
API: deepseek-v4-pro (DeepSeek-V4-Pro-0813)
$0.66–$1.32 per 1M input tokens (cache miss); $1.98–$3.96 per 1M output tokensPay as you go, deducted from a topped-up or granted balance
Listed prices: cache-hit input $0.022 off-peak or $0.044 at peak per 1M tokens; cache-miss input $0.66 or $1.32; output $1.98 or $3.96; no vision; concurrency limit 500. DeepSeek is phasing out V4-Pro: since 04:00 UTC on 14 September 2026 its requests route to V4.1-Flash at V4.1-Flash rates until V4.1-Pro launches.Official source 1 ↗ · Official source 2 ↗
Billing details
API charges are tokens multiplied by price and are deducted from a topped-up or granted balance, using the granted balance first. DeepSeek recommends topping up based on actual usage.Official source 1 ↗
Off-peak rates are half of peak rates. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday, excluding Chinese public holidays; all other hours, weekends and Chinese public holidays are off-peak.Official source 1 ↗
Images sent to the API are converted to tokens by size and billed with text tokens, up to 1024 tokens per image.Official source 1 ↗
DeepSeek states that prices may vary and that it can adjust them; the current prices took effect at 04:00 UTC on 10 September 2026.Official source 1 ↗ · Official source 2 ↗
Free access & limits
Chatting in the official DeepSeek app is free. DeepSeek's January 2025 app announcement says it has no ads and no in-app purchases. The checked pages do not state message or usage caps for the free app.Official source 1 ↗ · Official source 2 ↗
Best for & limitations
Asking questions in the app with web search or the Deep-Think reasoning mode turned on.Official source 1 ↗
Uploading a file to extract its text and work with it in a chat.Official source 1 ↗
Using DeepSeek as the backend model in coding agents such as Claude Code, GitHub Copilot or OpenCode without writing code.Official source 1 ↗
Reading text from screenshots or analysing charts through the API with deepseek-flash.Official source 1 ↗
Scheduling flexible API workloads in off-peak hours, when rates are 50% of peak rates.Official source 1 ↗
Answers can be inaccurate. DeepSeek's privacy policy says its models predict the most likely next words, which may not be factually accurate, and that you should not rely on the factual accuracy of outputs.Official source 1 ↗
Inputs can be used for training. DeepSeek collects text, voice, prompts, uploaded files and chat history, and uses personal data to train and improve its models. Depending on where you live, you may have the right to opt out of training use.Official source 1 ↗
Operated from China. The consumer services are provided and controlled by Hangzhou DeepSeek Artificial Intelligence Co., Ltd., registered in China.Official source 1 ↗
Vision only on deepseek-flash. Image input is supported on deepseek-flash only, and only in user messages; images in system or assistant messages return an error.Official source 1 ↗ · Official source 2 ↗
Desktop Harness on Apple silicon. The macOS version of DeepSeek Harness requires Apple silicon and macOS 13 or later.Official source 1 ↗
Submit a tool
