事情是这样，楼主今天正常使用codex，和同事一起用的 CPA 反代的账号，楼主觉得用起来很流畅，但是同事觉得用起来很慢。 
检查配置发现楼主和同事的 codex ~/.codex/config.toml 配置不同 
这是楼主的配置 
model_provider = "custom"
model = "gpt-5.6-sol"
model_reasoning_effort = "high"
disable_response_storage = true
personality = "pragmatic"

[model_providers.custom]
name = "custom"
wire_api = "responses"
requires_openai_auth = true
base_url = "https://xxx/v1"
supports_websockets = t...