[Bug]: Local Ollama models wait minutes without useful progress through OpenCode #8505

Platform

Windows

Installation source

Other

OpenBot version

0.68.0

Operating system and hardware

Windows 10.0.26200 (x64)

Agent runs on

Desktop app

Seen in

Desktop app

Providers

ChatGPT (Codex CLI)

Provider CLI version

0.167.2

Model

gpt-6-astra

Reasoning effort

high

Area

Agents and providers

Severity

Works, but incorrectly or slowly

What happened?

A short request to Qwen3.5 9B through OpenBot reported no response from OpenCode for over three minutes. Direct local chat responded, while a synthetic tool-enabled OpenCode test expanded a short arithmetic question to 5,320 input tokens and spent minutes processing the prompt on the CPU. Please expose prompt-processing progress and provide a lightweight local-chat configuration or guidance for CPU-based models.

Sent from the OpenBot app. An agent wrote this text, and the user approved it before it was sent.