github / github/copilot-cli

Bug with BYOK: reasoning effort not supported for model "glm-5.2:cloud"

Open
#4,012 2 comments 23 reactions 0 assignees View on GitHub
area:configuration area:models
Dominant language
Shell
Stars
11.2k
Forks
1.9k
Avg merge
14h 16m
Merged PRs (30d)
6

Description

### Describe the bug

I am experiencing an issue when using a custom BYOK configuration with Copilot CLI. Specifically, when I try to use the `--reasoning-effort max` flag, the CLI returns an error stating that the model does not support it, even though the configuration is otherwise valid.

### Affected version

v1.0.68

### Steps to reproduce the behavior

1. Run the following command:

COPILOT_PROVIDER_TYPE=openai COPILOT_PROVIDER_WIRE_API=responses COPILOT_PROVIDER_BASE_URL=https://ollama.com/v1 COPILOT_PROVIDER_API_KEY=$OLLAMA_API_KEY COPILOT_MODEL=glm-5.2:cloud COPILOT_PROVIDER_MAX_PROMPT_TOKENS=999424 COPILOT_PROVIDER_MAX_OUTPUT_TOKENS=131072 copilot --yolo --reasoning-effort max -p hi

2. Observe the error:
Error: Model "glm-5.2:cloud" does not support reasoning effort configuration (requested: "max").

### Expected behavior

The CLI should handle the reasoning effort flag gracefully or allow it for models that might support it, or at least provide a clearer error message indicating why it's not supported for this specific model configuration. Ideally, it should work as expected if the underlying API supports it.

### Additional context

I am confident I can use the command `COPILOT_PROVIDER_TYPE=openai COPILOT_PROVIDER_WIRE_API=responses COPILOT_PROVIDER_BASE_URL=https://ollama.com/v1 COPILOT_PROVIDER_API_KEY=$OLLAMA_API_KEY COPILOT_MODEL=glm-5.2:cloud COPILOT_PROVIDER_MAX_PROMPT_TOKENS=999424 COPILOT_PROVIDER_MAX_OUTPUT_TOKENS=131072 copilot --yolo --reasoning-effort max` without `-p` flag.

This issue only exists when using the `-p` flag.

Contributor guide

Open the contributing guide

Research direction

Start by running the reported command with and without the -p flag, using the listed BYOK environment variables, and compare how reasoning-effort validation is reached. Investigate the CLI entry path for -p and the model capability check. Done means the configuration is handled consistently, or the unsupported-model error clearly explains the limitation.

Written by the indexing model from the issue text.

Assessment

Tech stack
ollama, shell
Domain
api, cli
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
52/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.