GPT-5.6 Luna cannot be used with OpenAI/OpenAI Compatible provider (chat/completions + reasoning_effort)

[tustuntas]

Environment

• Cline: 4.1.2
• VS Code: 1.131.0 (user setup)
• OS: Windows 11 x64 10.0.22631
• Provider: OpenAI
• Also tested with: OpenAI Compatible
• API: Official OpenAI API (https://api.openai.com/v1)
• Model: gpt-5.6-luna

Description

My OpenAI account has access to gpt-5.6-luna. The model is returned by GET https://api.openai.com/v1/models:
json { "id": "gpt-5.6-luna", "object": "model" }
However, Cline only shows gpt-5.5 in the OpenAI provider model list.

As a workaround I switched to OpenAI Compatible and manually entered:

• Base URL: https://api.openai.com/v1
• Model: gpt-5.6-luna

The request reaches OpenAI, but always fails with:
json { "message": "400 Function tools with reasoning_effort are not supported for gpt-5.6-luna in /v1/chat/completions. To use function tools, use /v1/responses or set reasoning_effort to 'none'." }
Changing Reasoning Effort to None inside Cline does not change the behavior.

The same API key works correctly with direct OpenAI API calls outside Cline. This confirms the issue is in Cline's OpenAI integration, not the API key or account.

Expected behavior

Either:

• GPT-5.6 Luna should appear in the OpenAI model picker, or
• Cline should automatically use /v1/responses for GPT-5.6 models, or
• Cline should avoid sending reasoning_effort when using /v1/chat/completions with unsupported models.

Possible fix

Detect gpt-5.6-* models and route them through the Responses API instead of Chat Completions.

Additional information

• API key works and billing is active.
• gpt-5.6-luna is available from /v1/models.
• cline_models.json already contains entries for openai/gpt-5.6-luna, openai/gpt-5.6-terra, openai/gpt-5.6-sol.
• This appears to be a provider integration issue rather than an API access issue.

[linear-code]

CLINE-2815

[metrofun]

You can work around this without changing Cline by translating the request to Responses at the local AI gateway.

I verified this with swobuforge/swobu: Cline sends Chat Completions function tools plus:

reasoning_effort: "medium"

Swobu sends the upstream model a Responses request with the function tool and:

reasoning.effort: "medium"

No Chat-only reasoning_effort field reaches the Responses endpoint.

curl -fsSL https://swobu.com/install.sh | sh

Point Cline's OpenAI Compatible provider at Swobu and route the model to a Responses target.