Copilot: output_config.effort "max" is not supported by model claude-opus-4.6

[d4rky-pl]

Description

Seems that Copilot has changed the rules silently again and max effort is no longer supported (it worked yesterday 🤷)

Plugins

--

OpenCode version

1.4.6

Steps to reproduce

1. Use Claude Opus 4.6 from Copilot subscription
2. Select max model variant
3. Observe the following API error:

Bad Request: {"error":{"message":"output_config.effort \"max\" is not supported by model claude-opus-4.6; supported values: [low medium high]","code":"invalid_reasoning_effort"}}

Screenshot and/or share link

No response

Operating System

macOS 15.7.4, Docker

Terminal

iTerm2

[github-actions]

This issue might be a duplicate of or related to existing issues. Please check:

• fix: GitHub Copilot Claude models have wrong context limits and missing thinking variants #20317: GitHub Copilot Claude models have wrong context limits and missing thinking variants — this open issue also covers incorrect thinking/effort variants for Copilot Claude models, including claude-opus-4.6
• [Bug] Anthropic provider sends invalid API parameters (thinking.type: adaptive, output_config.effort) when variant is set #17876 (closed): Anthropic provider sends invalid output_config.effort when variant is set — a similar output_config.effort error, though on the native Anthropic provider rather than Copilot

[Onyelaudochukwuka]

Been running into this issue too with opencode v1.14.19

[ariefabrar]

me too, my solution is change the model and the variant/reasoning directly on opencode cli to high instead

[Onyelaudochukwuka]

That works for the parent agent, but subagents still spawn with the max variant, so even if the initial agent starts fine, it eventually hits the same error one level deeper when a subagent gets invoked.

[monotek]

I use oh my openagent and run into this too.

Currently downgraded to opencode 1.4.2 to be able to use it.

[JonasDoe]

Using omo as well, I adjusted the oh-my-opencode.json accordingly. But ofc. this should only be a workaround.

[Onyelaudochukwuka]

I also adjusted my oh-my-opencode.json config with reasoning level set across each agent, That didn't fix it for me

[erzz]

Its not directly related to OMO (though I spent hours thinking it was)

I replicate it without the plugin and a custom agent set to 4.6/high on the copilot provider - then the agent always just overrides it with 4.6/max which is not supported