Use Tokenify models in Dify. Tokenify serves DeepSeek V4 and GLM models through an OpenAI-compatible API at https://api.tokenify.dev/v1, billed per token, with credit that never expires.
- Create an account and an API key at https://www.tokenify.dev/app/signup.
- In Dify, open Settings → Model Provider, find Tokenify and click Add Model.
- Enter the model ID exactly as Tokenify lists it, for example
deepseek/deepseek-v4-pro, and your API key. - Optional: set the context size and the max-token cap, and turn on vision for models that accept images (
deepseek/deepseek-v4.1-flash,z-ai/glm-5.3-flash).
The current model list with prices is at https://www.tokenify.dev/models/ and from GET https://api.tokenify.dev/v1/models.
| Model ID | Notes |
|---|---|
deepseek/deepseek-v4.1-flash |
fast, accepts images |
deepseek/deepseek-v4-pro |
reasoning, agents and code |
deepseek/deepseek-v4-flash |
fast |
z-ai/glm-5.3-flash |
fast, accepts images |
z-ai/glm-5.2 |
reasoning |
- The endpoint is fixed to
https://api.tokenify.dev/v1; the plugin calls no other host. - Streaming and tool calling are supported.
- Reasoning models return their reasoning in
reasoning_content; reasoning tokens are billed at the output rate.
- A Tokenify API key (sign up at https://www.tokenify.dev/app/signup; usage is prepaid credit).
- Outbound HTTPS from the Dify plugin runtime to
api.tokenify.dev.
support@tokenify.dev · Docs: https://www.tokenify.dev/docs/ · Source: https://github.com/Graviti-AI/tokenify-dify-plugin