[DOCS]: Add guidance for GPT-5 unsupported parameter `stop` error in Dify

[l1cr0000-dev]

Type of Documentation Change

Error/Typo fix

Documentation Page URL or Path

https://docs.dify.ai/zh/learn/tutorials/workflow-101/lesson-08

Current Content

The current documentation does not clearly explain how to handle model compatibility issues when using newer OpenAI models such as GPT-5 in Dify.

When configuring GPT-5 as the model provider, Dify may send a stop parameter in the model request. However, GPT-5 does not support the stop parameter in this context, which causes the request to fail.

The error message is:

Failed to transform agent message: PluginInvokeError:
read llm model failed: request failed: [models] Bad Request Error, Error code: 400 -
{'error': {'message': "Unsupported parameter: 'stop' is not supported with this model.",
'type': 'invalid_request_error',
'param': 'stop',
'code': 'unsupported_parameter'}}

<img width="1470" height="956" alt="Image" src="https://github.com/user-attachments/assets/78339c18-c9e7-450d-9801-a87c8999007c" />

<img width="386" height="590" alt="Image" src="https://github.com/user-attachments/assets/ca9e4c06-ca16-43f4-b956-4cdef4945f63" />

### Suggested Changes

Please add documentation explaining model compatibility issues when using newer OpenAI models such as GPT-5 in Dify.

Suggested additions:

1. Clarify whether GPT-5 is currently fully supported by Dify's OpenAI model provider.
2. Explain that some newer OpenAI models may not support legacy parameters such as `stop`.
3. Add troubleshooting guidance for this error:

```text
Unsupported parameter: 'stop' is not supported with this model.

### Reason for Change

This change would help users understand why GPT-5 may fail in Dify even when GPT-4o works correctly in the same workflow.

Currently, when this error occurs, the Dify UI does not seem to provide a clear place to remove or disable the `stop` parameter manually. This makes it difficult for users to determine whether the issue is caused by their prompt, workflow configuration, model provider settings, or Dify's internal model request handling.

Adding documentation for this issue would reduce confusion, help users choose compatible models, and provide a clearer troubleshooting path for OpenAI model configuration problems.

### Code of Conduct

- [x] I agree to follow Dify's documentation [contribution guidelines](https://github.com/langgenius/dify/blob/0277a37fcad5ad86aeb239485c27fffd5cd90043/CONTRIBUTING.md)
- [x] I have searched for existing issues [search for existing issues](https://github.com/langgenius/dify/issues), including closed ones.
- [x] I confirm that I am using English to submit this report, otherwise it will be closed.
- [x] 【中文用户 & Non English User】请使用英语提交，否则会被关闭 ：）
- [x] Please do not modify this template :) and fill in all the required fields.

[dosubot]

Hi @l1cr0000-dev! I'm Dosu and I'm helping the Dify team.

This is a known behavior caused by a gap in the OpenAI provider plugin's parameter handling. Here's what's happening:

Root cause: Dify's OpenAI plugin does not currently filter out unsupported parameters before sending requests to the API. When a stop parameter is provided (either explicitly or by the workflow), the plugin adds it to the request regardless of whether the target model supports it [1]. GPT-5's model configuration YAML correctly declares that stop is not a supported parameter — only max_tokens, response_format, json_schema, reasoning_effort, verbosity, enable_stream, and service_tier are listed [2]. However, this information isn't used to filter parameters at runtime.

Why GPT-4o works but GPT-5 doesn't: GPT-4o's configuration includes stop as a supported parameter [3], so the API accepts it. GPT-5 (classified as a "thinking model" alongside the o-series) has a much more restricted parameter set [4].

Current workaround: Remove any explicit stop sequences from your node configuration. In the LLM node settings, check if you've set any stop sequences and clear them. Note that this may not fully resolve the issue if Dify is injecting stop internally as part of agent/workflow orchestration.

Fix status: The Dify team has been actively building parameter filtering infrastructure — a PR merged in May 2026 added filtering of preset parameters based on supportedParameterNames, and another PR from June 2026 extends capability-based filtering at the backend. The proper fix would extend this filtering to ensure the stop parameter (and other unsupported params) are stripped before API calls to models that don't support them.

This is more of a bug in the plugin's parameter handling than a documentation gap — the model YAML already correctly declares what's supported, but the runtime doesn't enforce it. I'd suggest also opening an issue on dify-official-plugins if one doesn't already exist, as that's where the OpenAI provider code lives.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.