LLM responses are cropped in "plan" mode. #24918

Description

OpenCode Go subscriber here. LLM generation often stops mid-message in plan mode, a few tokens into the final message. Thinking tokens are all generated.

A typical example will be as follow :

<think>
I am processing the user request, I need to ask 3 questions
</think>

I have 3 questions : 
1. What is 

As you can see the end of the message is cut. retrying, and even restarting OpenCode does not fix the issue. The issue shows on all opencode Go and opencode Zen models.

Plugins

No plugin

OpenCode version

1.14.29

Steps to reproduce

1. Carry a long-ish conversation in plan mode (architecturing a refactor for instance)
2. After a few (2-3 messages) every new message starts being cut before the end

Note : I regularly prompt OpenCode in French, which may or may not have an impact here

Screenshot and/or share link

Here is a screenshot of an instance of the issue :

Operating System

macOS 26.4.1 (25E253)

Terminal

Apple Terminal & vscode