Model Specifications

• Model: DeepSeek R1 Distill LLaMA 70B (bfloat16)
• Input Context: 131k tokens
• Generation Speed: ~20 tokens/second
• Live Web Search Capabilities
• Support: Available on Discord https://discord.gg/DDw9vS7g

Endpoint Description

This API provides access to the DeepSeek R1 language model, allowing you to generate responses to user messages with configurable parameters and real-time web search capabilities.

Authentication

• Requires API key in the request headers -> -H "x-billing-token: {your_api_token}"

Request Body Parameters

Required Parameters

| Parameter | Type | Description | Example | |-----------|------|-------------|---------| | messages | array | An array of message objects representing the conversation | [{"role": "user", "content": "Hello"}] |

Optional Parameters

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | max_tokens | integer | 2048 | Maximum number of tokens to generate in the response | | temperature | float | 0.7 | Controls randomness in response generation (0.0 - 1.0) | | top_p | float | 0.9 | Controls diversity of token selection (0.0 - 1.0) | | stream | boolean | false | Whether to stream the response | | search_web | boolean | false | Enable web search to enhance response |

Message Object Structure

{
  "role": "user" | "assistant" | "system",
  "content": "Message text here"
}

Roles

• user: Messages from the end-user
• assistant: Previous AI responses (optional, for context)
• system: System-level instructions (optional)

Example Request

model_name could be one of (depending on your active plan):
-> deepseek-r1-distill-llama-70b-websearch

curl -X POST "https://deadlock.p.nadles.com/{model_name}/completion" \
  -H "Content-Type: application/json" \
  -H "x-billing-token: {your_api_token}" \
  -d '{
    "messages": [
      {
        "role": "user",
        "content": "just respond with a short hi, nothing more"
      }
    ],
    "max_tokens": 2048,
    "temperature": 0.7,
    "top_p": 0.9,
    "stream": false,
    "search_web": false
  }'

Example Response

{
  "model_response": {
    "choices": [
      {
        "finish_reason": "stop",
        "index": 0,
        "logprobs": null,
        "message": {
          "content": "content text here",
          "name": null,
          "role": "assistant",
          "tool_calls": null
        }
      }
    ],
    "created": 1738346112,
    "id": "chatcmpl-f90e0bbff53d4b9390dc7e44bf67d044",
    "model": "deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
    "object": "chat.completion"
  },
  "summaries": [
    {
      "error": null,
      "status": 200,
      "summary": "summary text here",
      "url": "source_url_here"
    }
  ],
  "token_info": {
    "prompt_tokens": 431,
    "response_tokens": 814,
    "total_tokens": 1245
  }
}

Response Fields

| Field | Type | Description | |-------|------|-------------| | model_response.choices[].finish_reason | string | Reason for response completion | | model_response.choices[].index | integer | Index of the choice | | model_response.choices[].logprobs | object|null | Log probabilities of tokens | | model_response.choices[].message.content | string | Generated response text | | model_response.choices[].message.name | string|null | Name identifier | | model_response.choices[].message.role | string | Role of the message sender | | model_response.choices[].message.tool_calls | array|null | Tool call information | | model_response.created | integer | Timestamp of creation | | model_response.id | string | Unique response identifier | | model_response.model | string | Model name | | model_response.object | string | Object type | | summaries | array | Web search results | | summaries[].error | string|null | Error information | | summaries[].status | integer | HTTP status code | | summaries[].summary | string | Search result summary | | summaries[].url | string | Source URL | | token_info.prompt_tokens | integer | Number of tokens in prompt | | token_info.response_tokens | integer | Number of tokens in response | | token_info.total_tokens | integer | Total tokens used |

Pricing

• Pricing is based on token usage