Add qwen3.5-flash, qwen3.5-plus, qwen3-max model

[shilkazx]

Summary

Changes

Testing

• cargo clippy --workspace --all-targets -- -D warnings passes
• cargo test --workspace passes
• Live integration tested (if applicable)

Security

• No new unsafe code
• No secrets or API keys in diff
• User input validated at boundaries

[jaberjaber23]

Reviewed. Please provide a documentation link proving qwen3.5-flash, qwen3.5-plus, and qwen3-max exist in Alibaba's API. The model names don't match any known public Qwen releases. Also: qwen3.5-flash has a self-referencing alias (alias = model ID), and the relationship between new qwen3-max and existing qwen-max needs clarification.

[shilkazx]

Reviewed. Please provide a documentation link proving qwen3.5-flash, qwen3.5-plus, and qwen3-max exist in Alibaba's API. The model names don't match any known public Qwen releases. Also: qwen3.5-flash has a self-referencing alias (alias = model ID), and the relationship between new qwen3-max and existing qwen-max needs clarification.

OK. I'll paste each model's link here. Please note that the web page is written in Chinese, but it contains the API code example in Python.

1. qwen3.5-plus
2. qwen3.5-flash
3. qwen3-max

Qwen3-max is the next-generation model of qwen-max. Based on the official intro, qwen-max is a 2.5 generation model, and qwen3-max is a 3rd generation model.
The qwen-max model page is different from qwen3-max: qwen-max page

If you have any further questions, please let me know.

[shilkazx]

Reviewed. Please provide a documentation link proving qwen3.5-flash, qwen3.5-plus, and qwen3-max exist in Alibaba's API. The model names don't match any known public Qwen releases. Also: qwen3.5-flash has a self-referencing alias (alias = model ID), and the relationship between new qwen3-max and existing qwen-max needs clarification.

A further question is why not let a config file(like a toml/json) take responsibility for model spec(but provider spec should be retained in source code for security concerns).
The pros of config file is more flexible for users to configure some uncommon models from providers, and easier to extend/update the model list(i.e. reduce PR review workload; allow user use newly released model before openfang update).
Does openfang project concern something I don't realize?

[Jengro777]

https://github.com/librefang/librefang