AnyModelAnyModel
Начать
DeepSeek

DeepSeek V4 Flash Vision

DeepSeek V4 Flash that reads images, with the full 1M window.

DeepSeek V4 Flash Vision is the text+image variant of the V4 Flash line: the same cheap reasoner with image input and a 1M-token window, served by AnyModel over its own managed capacity. Pick it when a Flash-priced model has to look at screenshots, diagrams or scanned pages; ds/deepseek-v4-flash stays the text-only route.

Лучше всего для

Visual debuggingDocument extractionBudget multimodal

Сильные стороны

  • ▸Image understanding
  • ▸Контекст 1M токенов
  • ▸Low price

Использование DeepSeek V4 Flash Vision через AnyModel

Один OpenAI-совместимый эндпоинт — просто укажите id модели:

curl https://anymodel.org/v1/chat/completions \
  -H "Authorization: Bearer $ANYMODEL_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "ds/deepseek-v4-flash-vision",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
Установка Codex одной строкойПолучить ключ