DeepSeek: DeepSeek V4 Flash logo

MODELS

DeepSeek: DeepSeek V4 Flash

by DeepSeek

modellead-sourceopenrouter-modelsllmdeepseekdeepseek-v4-flashlong-contexttool-callingjson-outputopenai-compatibleanthropic-compatiblesource:api-docs.deepseek.comdeepseek-v4text-generation
Visit WebsiteDocumentation

Overview

DeepSeek-V4-Flash is a DeepSeek API model with 1M-token context, 384K maximum output, tool calls, JSON output, and OpenAI ChatCompletions and Anthropic-compatible access.

Details

DeepSeek-V4-Flash is listed by DeepSeek as the API model parameter `deepseek-v4-flash`. DeepSeek’s pricing documentation lists a 1M context length, 384K maximum output, tool calls, JSON output, and published token pricing. DeepSeek’s V4 Preview Release describes V4-Flash as a 284B-total/13B-active-parameter model available in chat and API, with Thinking and Non-Thinking modes, plus OpenAI ChatCompletions and Anthropic API support. The DeepSeek V4 model card identifies DeepSeek-V4-Flash as a text-modality model released April 24, 2026.

When to Use

Use when you need a DeepSeek-hosted text model with a 1M-token context window. Use when your application needs high maximum output length tool calls or JSON output through DeepSeek’s API. Use when you want to access a DeepSeek V4 model through OpenAI ChatCompletions-compatible or Anthropic-compatible interfaces.

Getting Started

  1. Review DeepSeek’s Models & Pricing page for current `deepseek-v4-flash` limits and token pricing.
  2. Use `deepseek-v4-flash` as the model parameter in DeepSeek API requests.
  3. Choose the OpenAI ChatCompletions-compatible interface or the Anthropic-compatible interface as documented by DeepSeek.
  4. Test your target context length
  5. JSON output
  6. tool-call behavior
  7. and Thinking/Non-Thinking mode behavior before production use.

Key Features

  • •1M-token context length listed in DeepSeek’s official pricing documentation and model card.
  • •384K maximum output listed for DeepSeek-V4-Flash.
  • •Supports tool calls and JSON output according to DeepSeek’s model pricing documentation.
  • •Available through OpenAI ChatCompletions and Anthropic-compatible interfaces according to DeepSeek’s V4 release notes and change log.
  • •DeepSeek describes V4-Flash as a 284B-total/13B-active-parameter model.
  • •Model card identifies DeepSeek-V4-Flash as a text-modality model released April 24
  • •2026.

Capabilities

  • •text generation
  • •long-context processing
  • •tool calling
  • •JSON output
  • •OpenAI-compatible chat completions
  • •Anthropic-compatible API access

Last updated May 31, 2026