Error: GLM Flash - Unknown error

[scimsoft]

Error Details

Model: GLM Flash
Provider: openai
Status Code: N/A

Error Output

Invalid tool name

Additional Context
Please add any additional context about the error here

[sestinj]

Related issues involving GLM models and tool calling errors:

• [Bug] OpenAI provider uses legacy /completions endpoint for autocomplete with GLM-4 #8419 - OpenAI provider uses legacy /completions endpoint for autocomplete with GLM-4
• edit_existing_file may need an improved description to help GLM-4.6 use it correctly all the time #8189 - edit_existing_file may need an improved description to help GLM-4.6 use it correctly all the time
• Error: OpenAI GPT-4.1 mini - 400 #9785 - Error: OpenAI GPT-4.1 mini - 400 (Invalid tool_calls function.name: empty string)
• Error: GPT-4o Mini - 400 #9784 - Error: GPT-4o Mini - 400 (Invalid tool_calls function.name: empty string)

[RomneyDa]

@scimsoft does this happen when using specific MCP servers? If so could you share which tool names are causing the issue?

[RomneyDa]

Fixed by #11792.

What was happening: Your model wasn't recognized in Continue's native tool support list (PROVIDER_TOOL_SUPPORT), so it fell back to a text-based tool calling format injected into the system message. When the model didn't follow that format exactly, the parser threw "Invalid tool name". This PR adds your model family to the allowlist so it uses native tool calling instead.

Workaround (if you're on an older version): In your model config, set "capabilities": { "tools": true } to force native tool calling.