From dc7e52874e281fc70501e554121e771feb9cbd1e Mon Sep 17 00:00:00 2001
From: teknium1 <127238744+teknium1@users.noreply.github.com>
Date: Tue, 15 Sep 2026 12:39:19 -0700
Subject: [PATCH] docs(agent): record why the Z.AI vision default is
 glm-5.3-flash
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Pinned vision ids rot silently (glm-5v-turbo was retired from the Coding Plan endpoints
while still valid on pay-as-you-go), so name the data behind the new pin — the only
image-capable GLM id on every Z.AI surface — and why the pin cannot simply be dropped in
favour of ProviderProfile.default_vision_model() (ZaiProfile returns None, which would route
vision to a text-only chat model).

Co-authored-by: POWERFULMOVES <142271328+POWERFULMOVES@users.noreply.github.com>
---
 agent/auxiliary_client.py | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/agent/auxiliary_client.py b/agent/auxiliary_client.py
index f0220ba541..1f3395ee2f 100644
--- a/agent/auxiliary_client.py
+++ b/agent/auxiliary_client.py
@@ -789,7 +789,11 @@ def _task_prefers_fast_model(task: Optional[str]) -> bool:
         _get_auxiliary_task_config(task).get("prefer_fast_model"), default=False)
 
 
-# Dedicated vision models for direct providers whose main chat model differs.
+# Dedicated vision models for direct providers whose main chat model differs. zai: glm-5.3-flash
+# is the only image-capable GLM id served on every Z.AI surface (pay-as-you-go and Coding Plan,
+# global and CN); the former glm-5v-turbo pin 404s / 1211 "Unknown Model" on the coding endpoints
+# (#111429). ZaiProfile has no default_vision_model(), so dropping the pin would route vision to
+# the user's text-only chat model and skip Z.AI entirely.
 _PROVIDER_VISION_MODELS: Dict[str, str] = {"xiaomi": "mimo-v2.5", "zai": "glm-5.3-flash"}