DeepSeek Platform
** – API and playground
Last verified:
What is DeepSeek Platform?
The DeepSeek Platform is an API service that provides programmatic access to DeepSeek's advanced AI models, including deepseek-v4-flash, deepseek-v4-pro, deepseek-chat, and deepseek-reasoner. The platform offers an OpenAI/Anthropic-compatible API format, enabling developers to integrate powerful AI capabilities like natural language processing, reasoning, code generation, and chat into their applications with minimal setup changes.
Key features include JSON Output for structured responses, Chat Prefix Completion for specifying assistant message prefixes, FIM (Fill-In-the-Middle) Completion for code and story completion, Function Calling support, streaming responses, and a Beta API with extended max_tokens up to 8192. The platform supports multiple programming languages through SDKs for Python, Node.js, and curl, and integrates with agent tools like Claude Code, GitHub Copilot, and OpenCode.
The platform is designed for developers, startups, enterprises, and AI researchers who need to embed AI capabilities into applications. It serves use cases including chatbots, coding assistants, automation tools, research prototyping, and AI-driven products. The pay-as-you-go pricing model makes it accessible for both individual developers testing prototypes and enterprises scaling production workloads.
DeepSeek offers both thinking mode (reasoning-intensive tasks) and non-thinking mode (faster responses) through its models. The platform supports a 128K context window for deepseek-chat and up to 1M context for deepseek-v4-pro, with maximum output tokens ranging from 4K to 384K depending on the model.
DeepSeek Platform pricing
Pricing model: Freemium
DeepSeek uses pay-as-you-go pricing with no classic subscription tiers. New users receive 5 million free tokens on sign-up with no credit card required. The free tier allows 1-3 million tokens monthly, 10-30 requests per minute, 1-3 concurrent requests, and file uploads up to 10MB. Commercial use requires upgrade to paid plan. Pricing per 1M tokens: deepseek-chat costs $0.07 (cache hit) / $0.27 (cache miss) input and $1.10 output; deepseek-reasoner costs $0.14 (cache hit) / $0.55 (cache miss) input and $2.19 output. DeepSeek V4 Flash costs $0.0028/M cached input, $0.14/M cache-miss input, $0.28/M output. DeepSeek V4 Pro costs $0.003625/M cached input, $0.435/M cache-miss input, $0.87/M output. V4 Pro has a 75% discount until May 31, 2026. The consumer chat app at deepseek.com is completely free with no paywall.
DeepSeek Platform pros
- OpenAI/Anthropic API compatible - easy migration from existing services
- Free tier with 5 million tokens for new users on sign-up
- Extremely competitive pricing - up to 200x cheaper than some competitors
- DeepSeek-v4-pro offers 1M token context window
- JSON Output feature enforces valid JSON format for structured data
- FIM Completion supports code completion and story completion
- Chat Prefix Completion allows custom assistant message prefixes
- Streaming responses supported for real-time text display
- Beta API extends max_tokens from 4K to 8192
- Function Calling support for tool integration
- Integrates with Claude Code, GitHub Copilot, OpenCode out of the box
- Python, Node.js, and curl SDK examples provided
- Both thinking mode and non-thinking mode available
- No rate limits enforced on API - serves every request possible
- 75% discount on V4-Pro available until May 31, 2026
DeepSeek Platform cons
- deepseek-chat and deepseek-reasoner deprecated on July 24, 2026
- Free plan restricted to non-commercial use only
- Rate limits are dynamic and cannot be increased per account
- Beta features require setting base_url to https://api.deepseek.com/beta
- JSON Output may occasionally return empty content
- Maximum output tokens limited to 8K for most models
- Only email registration supported in some regions
- No tiered subscription plans - only pay-as-you-go pricing
Frequently asked questions about DeepSeek Platform
How do I get a DeepSeek API key?
Visit platform.deepseek.com, create an account or sign in using email registration, navigate to the API Keys section on the left sidebar, click 'Create New API Key', name your key (e.g., 'test key'), click 'Create API Key', then copy and securely store the key for use in your applications.
What is the base URL for the DeepSeek API?
The base URL for the OpenAI-compatible API is https://api.deepseek.com. For the Anthropic-compatible API, use https://api.deepseek.com/anthropic. For Beta features like Chat Prefix Completion, FIM Completion, and extended max_tokens, set base_url to https://api.deepseek.com/beta.
Which models are available on the DeepSeek API?
Available models include deepseek-v4-flash (for high-volume agent traffic), deepseek-v4-pro (for harder coding and reasoning with 1M context), deepseek-chat (deprecated July 24, 2026, corresponds to non-thinking mode of v4-flash), and deepseek-reasoner (deprecated July 24, 2026, corresponds to thinking mode of v4-flash).
How do I enable JSON Output mode?
Set the response_format parameter to {'type': 'json_object'}, include the word 'json' in your system or user prompt with an example of the desired JSON format, and set max_tokens appropriately to prevent truncation. The API enforces valid JSON format string output.
What is the difference between thinking mode and non-thinking mode?
Thinking mode (enabled via thinking: {'type': 'enabled'} and reasoning_effort: 'high') produces more reasoning-intensive responses suitable for complex tasks. Non-thinking mode provides faster responses for simpler tasks. deepseek-reasoner corresponded to thinking mode and deepseek-chat to non-thinking mode before deprecation.
How do I use streaming responses?
Set the stream parameter to true in your API request. The response will be sent in chunks, allowing you to display text as it is generated. The streaming response format is slightly different and requires parsing the chunks.
What are the rate limits for the DeepSeek API?
Rate limits are dynamically adjusted based on real-time traffic pressure and each account's short-term historical usage. The platform temporarily does not support increasing the dynamic rate limit for individual accounts. Free accounts typically get 10-30 requests per minute with 1-3 concurrent requests.
Can I use DeepSeek for commercial applications?
No, the free plan is restricted to personal, academic, or non-commercial projects only. Commercial use requires upgrading to a paid plan. The pay-as-you-go pricing model permits commercial deployment for paid accounts.
How do I view my API usage by key?
Go to the Usage page, select the relevant month, click Export, download and unzip the usage data package. You will find two CSV files. The file named 'amount' contains usage details broken down by API Key.
What is FIM (Fill-In-the-Middle) Completion?
FIM Completion allows users to provide custom prefixes and/or suffixes for the model to complete content in the middle. It is commonly used for story completion and code completion scenarios. The FIM Completion API is charged the same as Chat Completion and requires setting base_url to https://api.deepseek.com/beta to enable.