Dify: OpenAI-API-compatible provider

Dify's official OpenAI-API-compatible provider adds custom models one by one. Field names below are from the official plugin.

Replace YOUR-ENDPOINT with the endpoint shown on your API Key page after you sign up, and YOUR-KEY with your key. One key works for every model on 666 AI — Claude, GPT, Gemini, DeepSeek, Qwen and 200+ more.

Steps

  1. Integrations → Model Provider → install OpenAI-API-compatible (from the Marketplace if needed). Only workspace owners and admins can do this.
  2. On the provider card click Add Model, type LLM.
  3. Model Name: e.g. anthropic/claude-sonnet-5.5.
  4. API Key: YOUR-KEY. API Base URL: https://YOUR-ENDPOINT/v1.
  5. Completion mode: Chat. Model context size: the model's real value (e.g. 200000) — the default 4096 is far too small.
  6. For agents and tools, set Function Call Type to Tool Call; turn on Vision if the model supports images.
  7. Save — Dify makes a test call before saving.

Embeddings

Add Model → type Text Embedding, Model Name openai/text-embedding-3-small, the same API Base URL with /v1.

API Type

Keep Chat Completions API. Claude models on 666 AI are not served on the Responses API.

FAQ

Answers are cut off or the context is lost

Raise Model context size and Upper bound for max tokens to the model's real limits.

Should I remove /v1 for embeddings?

No. Keep https://YOUR-ENDPOINT/v1 for every model type.

Create a free accountAll modelsDeveloper guides