# DeepSeek V4 Flash - Core.Today AI API

> DeepSeek's fast, low-cost V4 model (checkpoint V4-Flash-0731). Thinking mode is on by default, with a 1M-token context window and heavily discounted cache reads. Replaces the retired deepseek-chat / deepseek-reasoner.

- **Provider**: DeepSeek
- **Model ID**: deepseek-v4-flash
- **Category**: LLM
- **Credits**: 0.4 per 1K tokens (avg)
- **Speed**: Fast
- **Quality**: High

## Model Specifications
- **Context Window**: 1.0M tokens
- **Max Output**: 384K tokens
- **Training Cutoff**: 2025-05
- **Supported Formats**: text, json, markdown
- **Compatible SDK**: OpenAI

### Capabilities
- Function Calling
- Streaming
- JSON Mode
- System Prompt

### Token Pricing (per 1M tokens)
- **Input Tokens**: 260 credits ($0.17)
- **Output Tokens**: 520 credits ($0.35)
- **Cached Tokens**: 5.2 credits ($0.00)

## Features
- Best-in-class price/performance
- 1M context window, up to 384K output tokens
- Thinking mode on by default (switchable per request)
- OpenAI-compatible API (chat completions)
- Context cache hits billed at ~98% discount

## Use Cases
- High-volume chat and summarization at low cost
- Long-document analysis with the 1M context
- Migration target for retired deepseek-chat / deepseek-reasoner

## API Endpoint

Base URL: https://api.core.today
Endpoint: POST /llm/deepseek/v1/chat/completions

## Authentication
Header: Authorization: Bearer YOUR_API_KEY

Note: LLM endpoints use OpenAI-compatible format with Authorization Bearer token.

## Input Parameters

### Required
- **model**: string - Model ID, e.g. "deepseek-v4-flash".
- **messages**: array - Chat messages in OpenAI format (system/user/assistant roles).

### Optional
- **temperature**: number (default: 0.7) - Sampling temperature (0-2).
- **max_tokens**: integer (default: 2048) - Maximum completion tokens.
- **stream**: boolean (default: false) - Stream the response as server-sent events.

## Examples

### Chat completion
OpenAI-compatible chat completion request
```json
{
  "model": "deepseek-v4-flash",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in 3 sentences."
    }
  ],
  "max_tokens": 1024
}
```

## Response Format
```json
{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Response text here"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 100,
    "completion_tokens": 50,
    "total_tokens": 150
  }
}
```

## Tips
- Works with the OpenAI SDK - set base_url to https://ai.api.core.today/llm/deepseek/v1
- Thinking mode is on by default - disable it per request for latency-sensitive calls
- Cache hits are billed at ~98% off - reuse long system prompts and shared prefixes
- Anthropic-format endpoint is not supported through the gateway - use the OpenAI format

## Documentation
https://api-docs.deepseek.com/