DeepSeek-V4-Flash-Vision-Exp Release Notes

DeepSeek has released DeepSeek-V4-Flash-Vision-Exp, an experimental multimodal model available via the DeepSeek API. This model enables the integration of visual understanding with text-based reasoning and tool use, significantly improving multimodal agent performance to levels approaching Opus-4.8.

Multimodal Performance and Capabilities

DeepSeek-V4-Flash-Vision-Exp maintains the same text capabilities as the standard DeepSeek-V4-Flash model, including its reasoning, world knowledge, and agentic abilities. The primary advancement is in its multimodal capabilities, where it demonstrates a significant performance leap over V4-Flash on multimodal agent benchmarks, bringing its performance close to that of Opus-4.8.

Agentic Workflows and Visual Understanding

The model is designed to work seamlessly with agent frameworks. By combining visual understanding with a diverse set of tools, DeepSeek-V4-Flash-Vision-Exp allows for more practical and complex workflows that require the model to interpret visual data to execute tasks.

API Integration and Implementation

Developers can access the model using the identifier model='deepseek-v4-flash-vision-exp'. To facilitate immediate adoption, DeepSeek Harness 0.1.1 was released concurrently with the model to provide out-of-the-box support.

Input and Billing

  • Input Formats: The model supports mixed text and image inputs. Images can be provided via external URLs, base64 encoding, or the Files API.
  • Billing: Images are tokenized for billing purposes, with each image consuming up to 384 tokens. Pricing for these tokens is aligned with V4-Flash pricing.
  • Supported Endpoints: The model supports Chat Completions, Messages, and Responses.

Files API Launch

Alongside the model release, DeepSeek has launched a free-to-use Files API. This API allows users to upload an image once and reference it via a file_id in subsequent requests. This architecture reduces request bandwidth and eliminates the need to re-upload the same image across multiple API calls.

Sources

Related

  • Dispatch
    DeepSeek-V4 Preview Release

    DeepSeek has released the open-source DeepSeek-V4 preview, introducing two model variants—Pro and Flash—featuring a 1M token context window and state-of-the-art agentic coding capabilities.

  • Dispatch
    DeepSeek-V4-Flash Update

    DeepSeek has released a public beta update for DeepSeek-V4-Flash, significantly enhancing agent capabilities and benchmark performance through re-post-training while maintaining the original model architecture.

  • Dispatch
    DeepSeek-V4-Pro GA Release

    DeepSeek has released DeepSeek-V4-Pro, featuring enhanced agent capabilities, flexible reasoning effort settings, and native OpenAI Responses API support.

  • Dispatch
    DeepSeek v4 Flash Vision Experimental Model – API Guide and Community Insights

    DeepSeek’s deepseek‑v4‑flash‑vision‑exp model adds image input via OpenAI‑compatible APIs, supporting JPEG/PNG/GIF/WebP with flexible upload methods, token‑based pricing, and detailed limits.

  • Dispatch
    DeepSeek Introduces Vision Capabilities

    DeepSeek has integrated vision capabilities into its chat interface, enabling the model to understand and describe image content rather than just performing text extraction.