> ## Documentation Index > Fetch the complete documentation index at: https://docs.evolink.ai/llms.txt > Use this file to discover all available pages before exploring further. # DeepSeek V4 - Anthropic-Compatible API > - Invoke DeepSeek V4 models using the Anthropic Messages protocol - Supports `deepseek-v4-flash` / `deepseek-v4-pro` - Request / response structures aligned with the Anthropic API - **Plain text conversation** (image / document content types are not yet supported) - **System prompts**: Passed via the top-level `system` field - **Thinking mode**: Controlled by the `thinking` object; thinking content is returned as a `content[type=thinking]` block - **Streaming output**: SSE event stream - **Tool calling**: Compatible with Anthropic `tool_use` / `tool_result` flow **BaseURL**: The default BaseURL is `https://direct.evolink.ai`, which has better support for text models and long-lived connections. `https://api.evolink.ai` is the primary endpoint for multimodal services and serves as a fallback address for text models. ## OpenAPI ````yaml /en/api-manual/language-series/deepseek-v4/deepseek-v4-messages.json POST /v1/messages openapi: 3.1.0 info: title: DeepSeek V4 Anthropic-Compatible API description: >- Invoke the DeepSeek V4 series (`deepseek-v4-flash` / `deepseek-v4-pro`) via the Anthropic Messages protocol. **Compatibility notes**: - Path: `/v1/messages` (Anthropic standard path) - Request / response structures aligned with the Anthropic Messages API - Supported fields: `model` `max_tokens` (required) `messages` `system` `temperature` `top_p` `stop_sequences` `stream` `thinking` `tools` `tool_choice` `output_config` - **Unsupported fields**: `top_k`, `container`, `mcp_servers`, `metadata`, `service_tier`, `cache_control` - **Unsupported content types**: images (`image`), documents (`document`), search results, `redacted_thinking`, `server_tool_use` **Model capabilities**: - 1M token context, max output 384K tokens - Pro has `thinking` enabled by default; responses include a `thinking` content block **Billing tiers (UC/1K tokens, EvoLink internal unit)**: | Model | Input cache hit | Input cache miss | Output | | --- | --- | --- | --- | | deepseek-v4-flash | 20 | 100 | 200 | | deepseek-v4-pro | 100 | 1200 | 2400 | license: name: MIT version: 1.0.0 servers: - url: https://direct.evolink.ai description: Production (recommended) - url: https://api.evolink.ai description: Alternative URL security: - bearerAuth: [] tags: - name: Messages description: Anthropic Messages protocol endpoints paths: /v1/messages: post: tags: - Messages summary: DeepSeek V4 Messages API (Anthropic-Compatible) description: >- - Invoke DeepSeek V4 models using the Anthropic Messages protocol - Supports `deepseek-v4-flash` / `deepseek-v4-pro` - Request / response structures aligned with the Anthropic API - **Plain text conversation** (image / document content types are not yet supported) - **System prompts**: Passed via the top-level `system` field - **Thinking mode**: Controlled by the `thinking` object; thinking content is returned as a `content[type=thinking]` block - **Streaming output**: SSE event stream - **Tool calling**: Compatible with Anthropic `tool_use` / `tool_result` flow operationId: createMessageDeepSeekV4 requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/CreateMessageRequest' examples: simple: summary: Minimal runnable request value: model: deepseek-v4-flash max_tokens: 1024 messages: - role: user content: Hello, world system_prompt: summary: With system prompt + multi-turn value: model: deepseek-v4-pro max_tokens: 2048 system: You are a senior English technical editor. messages: - role: user content: Describe DeepSeek V4 in three sentences. thinking: summary: Explicitly configure thinking mode value: model: deepseek-v4-pro max_tokens: 4096 thinking: type: enabled output_config: effort: high messages: - role: user content: Prove Euler's identity e^(iπ) + 1 = 0 disable_thinking: summary: Disable thinking mode value: model: deepseek-v4-pro max_tokens: 512 thinking: type: disabled messages: - role: user content: 'In one sentence: what is the capital of Japan?' tool_use: summary: Tool calling (Anthropic tool_use style) value: model: deepseek-v4-pro max_tokens: 2048 messages: - role: user content: Query Shanghai's weather and tell me tools: - name: get_weather description: Query the weather for a specified city input_schema: type: object properties: city: type: string description: City name required: - city tool_choice: type: auto streaming: summary: Streaming output (SSE) value: model: deepseek-v4-flash max_tokens: 1024 stream: true messages: - role: user content: Write a short poem about spring responses: '200': description: Message object content: application/json: schema: $ref: '#/components/schemas/MessageResponse' examples: thinking_enabled: summary: thinking enabled (default, includes thinking content block) value: id: 53ee6690-e14a-4e6b-890b-a135100d51c7 type: message role: assistant model: deepseek-v4-flash content: - type: thinking thinking: >- The user is asking about Japan's capital — a basic geography question. The answer is Tokyo, just give it directly. signature: 53ee6690-e14a-4e6b-890b-a135100d51c7 - type: text text: The capital of Japan is **Tokyo**. stop_reason: end_turn stop_sequence: null usage: input_tokens: 7 cache_creation_input_tokens: 0 cache_read_input_tokens: 0 output_tokens: 77 service_tier: standard thinking_disabled: summary: thinking disabled (text block only) value: id: a42c8fa2-e1b7-4cd3-9c48-71d2f5c6a8e0 type: message role: assistant model: deepseek-v4-flash content: - type: text text: The capital of Japan is Tokyo. stop_reason: end_turn stop_sequence: null usage: input_tokens: 7 cache_creation_input_tokens: 0 cache_read_input_tokens: 0 output_tokens: 9 service_tier: standard tool_use: summary: Triggers a tool call (stop_reason=tool_use) value: id: b61d9e03-3a78-4b95-8612-54e7f2a9c1d3 type: message role: assistant model: deepseek-v4-pro content: - type: thinking thinking: >- The user wants to check Beijing's weather. I need to call the get_weather tool with parameter Beijing. signature: b61d9e03-3a78-4b95-8612-54e7f2a9c1d3 - type: text text: Sure, I'll look up the weather in Beijing for you. - type: tool_use id: toolu_01abc123xyz name: get_weather input: city: Beijing stop_reason: tool_use stop_sequence: null usage: input_tokens: 35 cache_creation_input_tokens: 0 cache_read_input_tokens: 0 output_tokens: 68 service_tier: standard '400': description: Invalid request parameters content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: message: Invalid request type: invalid_request_error request_id: req_xxx type: error '401': description: Authentication error content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: message: Authentication error type: authentication_error type: error '402': description: Insufficient quota content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: message: Insufficient quota type: billing_error type: error fallback_suggestion: https://evolink.ai/dashboard/credits '403': description: Permission error content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: message: Permission denied type: permission_error type: error '404': description: Model or resource not found content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: message: Model not found type: not_found_error type: error '429': description: Rate limited content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: message: Rate limited type: rate_limit_error type: error '500': description: Internal server error content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' '502': description: Gateway error content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' '503': description: Service temporarily unavailable content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' components: schemas: CreateMessageRequest: type: object required: - model - max_tokens - messages properties: model: type: string description: >- Model to invoke - `deepseek-v4-flash`: Fast general-purpose - `deepseek-v4-pro`: Deep reasoning **Tip**: Both models **have thinking enabled by default**, and responses always contain a `type="thinking"` content block; set `thinking.type="disabled"` explicitly to turn it off. An unspecified or unsupported model is automatically mapped to `deepseek-v4-flash`. enum: - deepseek-v4-flash - deepseek-v4-pro default: deepseek-v4-flash example: deepseek-v4-flash max_tokens: type: integer description: |- Maximum number of tokens to generate (**required**) **Notes**: - The V4 series can reach up to **384,000** - Tokens produced by thinking also count toward the max_tokens limit minimum: 1 maximum: 384000 example: 1024 messages: type: array description: >- List of conversation messages, alternating between user / assistant turns **Notes**: - At least one message must be included - The final message is typically `role=user` - `image` / `document` content types are not yet supported items: $ref: '#/components/schemas/InputMessage' minItems: 1 system: description: >- System prompt, used to define the AI's role and behavior **Notes**: - Accepts a string or an array of strings - Unlike the `system` message on the OpenAI endpoint, the Anthropic endpoint uses a top-level `system` field oneOf: - type: string example: You are a helpful assistant. - type: array items: type: object properties: type: type: string enum: - text text: type: string temperature: type: number description: >- Sampling temperature **Notes**: - Range `[0.0, 2.0]` - Default 1; higher values are more diverse, lower values are more deterministic minimum: 0 maximum: 2 default: 1 example: 1 top_p: type: number description: |- Nucleus sampling threshold **Notes**: - Range `[0, 1]` - Do not adjust temperature and top_p simultaneously minimum: 0 maximum: 1 default: 1 example: 1 stop_sequences: type: array description: |- Custom stop sequences **Notes**: - Generation stops when the model encounters any of these strings - Up to 4 entries (following Anthropic's specification) items: type: string maxItems: 4 stream: type: boolean description: |- Whether to return via SSE streaming - `true`: Server-Sent Events streaming return - `false`: Return the complete response all at once (default) default: false example: false thinking: type: object description: >- Thinking mode control (V4) **Notes**: - **Enabled by default on both models** (`type=enabled`) - When enabled, the response `content` array includes a `type="thinking"` reasoning block (billed as output tokens) - **Note**: The API **ignores** Anthropic's native `budget_tokens` field; use `output_config.effort` to control depth - In multi-turn conversations, simply place the thinking block from the previous response back into the assistant `content` array verbatim (the Anthropic protocol is more lenient — missing thinking will not cause an error, but preserving the signature helps context consistency) properties: type: type: string enum: - enabled - disabled description: |- - `enabled`: Enable deep thinking - `disabled`: Disable deep thinking default: enabled budget_tokens: type: integer description: >- **Ignored** — DeepSeek does not use Anthropic's budget_tokens; use `output_config.effort` instead output_config: type: object description: |- Output configuration (V4 extension) **Notes**: DeepSeek only supports the `effort` field properties: effort: type: string description: |- Reasoning effort level - `low`: Low effort, faster response - `medium`: Medium effort (default) - `high`: High effort, deeper reasoning enum: - low - medium - high default: medium tools: type: array description: |- List of tool definitions **Notes**: - Follows the Anthropic tool definition specification - `input_schema` uses a JSON Schema object items: $ref: '#/components/schemas/Tool' tool_choice: type: object description: >- Controls tool-calling behavior **Available types**: - `auto`: Model decides automatically (default when tools are provided) - `any`: Must call some tool (without specifying which) - `tool`: Must call the specified `name` - `none`: Forbid tool calls properties: type: type: string enum: - auto - any - tool - none name: type: string description: Tool name specified when `type="tool"` disable_parallel_tool_use: type: boolean description: Disable parallel tool calls (standard Anthropic field) MessageResponse: type: object description: Anthropic-style message response properties: id: type: string description: Unique message ID type: type: string enum: - message description: Response object type role: type: string enum: - assistant model: type: string description: Model actually used example: deepseek-v4-pro content: type: array description: |- List of response content blocks **Possible block types**: - `thinking`: Reasoning process (only when thinking is enabled) - `text`: Final answer text - `tool_use`: Tool call initiated by the model items: $ref: '#/components/schemas/OutputContentBlock' stop_reason: type: string description: |- Stop reason - `end_turn`: Natural completion - `max_tokens`: Reached the max_tokens limit - `stop_sequence`: Hit a stop_sequences entry - `tool_use`: Model triggered a tool call enum: - end_turn - max_tokens - stop_sequence - tool_use stop_sequence: type: - string - 'null' description: >- The specific sequence hit when stop_reason=`stop_sequence`; otherwise null usage: $ref: '#/components/schemas/AnthropicUsage' ErrorResponse: type: object properties: type: type: string enum: - error error: type: object properties: type: type: string description: >- Error type (e.g. invalid_request_error / authentication_error / billing_error, etc.) message: type: string description: Error description request_id: type: string description: Request trace ID InputMessage: type: object required: - role - content properties: role: type: string enum: - user - assistant description: >- Message role - `user`: User message (also used to return `tool_result` blocks) - `assistant`: Historical assistant reply (may contain `text` / `thinking` / `tool_use` blocks) ⚠️ **Does not accept `system`**: system prompts must go through the top-level `system` field; `role="system"` in messages will be rejected (400 unknown variant). content: description: >- Message content **Notes**: - Pass a string directly for plain text - Pass an array of content blocks (`text` / `tool_use` / `tool_result`) for structured content - **Not supported**: multimodal types such as `image` / `document` oneOf: - type: string - type: array items: $ref: '#/components/schemas/ContentBlock' Tool: type: object required: - name - input_schema properties: name: type: string description: |- Tool name **Notes**: - Only `a-zA-Z0-9_-` allowed - Max 64 characters description: type: string description: >- Description of the tool's functionality, helping the model decide when to call it input_schema: type: object description: |- JSON Schema object describing the tool's input parameters **Notes**: - `type` must be `object` - Should declare `properties` and `required` OutputContentBlock: type: object description: Content block within the response properties: type: type: string enum: - text - thinking - tool_use text: type: string description: Text when type=`text` thinking: type: string description: Reasoning-process text when type=`thinking` signature: type: string description: >- Integrity signature when type=`thinking` (Anthropic specification, used to verify the reasoning was not tampered with) id: type: string description: Tool call ID when type=`tool_use` name: type: string description: Tool name when type=`tool_use` input: type: object description: JSON input generated by the model when type=`tool_use` AnthropicUsage: type: object description: Token usage statistics (Anthropic specification) properties: input_tokens: type: integer description: Number of input tokens (portion not hitting cache) example: 10 output_tokens: type: integer description: Number of output tokens (including thinking) example: 30 cache_creation_input_tokens: type: integer description: >- Number of input tokens for cache creation (the current DeepSeek Anthropic endpoint does not perform cache writes, so this value is fixed at 0) example: 0 cache_read_input_tokens: type: integer description: >- Number of input tokens that hit the cache **Notes**: Billed at the cache-hit rate (Flash 20 UC/1K, Pro 100 UC/1K) example: 0 service_tier: type: string description: Service tier (Anthropic specification field) example: standard ContentBlock: type: object description: |- Message content block **Supported types**: - `text`: Text fragment - `tool_use`: Assistant-initiated tool call - `tool_result`: User-returned tool execution result properties: type: type: string enum: - text - tool_use - tool_result text: type: string description: Text content when type=`text` id: type: string description: Tool call ID (required for tool_use / tool_result) name: type: string description: Tool name (required for tool_use) input: type: object description: Tool input arguments (for tool_use, JSON object) tool_use_id: type: string description: >- Corresponding tool call ID (required for tool_result, matches tool_use.id) content: description: >- Tool execution result (tool_result); string or array of content blocks oneOf: - type: string - type: array items: type: object securitySchemes: bearerAuth: type: http scheme: bearer description: >- ##All APIs require Bearer Token authentication## **Get API Key**: Visit the [API Key Management Page](https://evolink.ai/dashboard/keys) to obtain your API Key **Add to request header**: ``` Authorization: Bearer YOUR_API_KEY ``` **Note**: Although Anthropic's native API uses the `x-api-key` header, EvoLink's `/v1/messages` uniformly uses Bearer Token authentication. ````