chore: bump GLM-5.1 max_tokens to 49152

Raise max_tokens from 32768 to 49152 for zai-org/GLM-5.1 and
zai-org/GLM-5.1-FP8 in prod and dev. Providers (Together, Fireworks AI,
zai-org) advertise 202,752-token context on the HF router, so the
larger output budget sits well within the context window.

https://claude.ai/code/session_01AjLGLnaXowm91ymkX42wmN

Author: claude

chart/env/dev.yaml

@@ -79,8 +79,8 @@ envVars:
   PUBLIC_LLM_ROUTER_ALIAS_ID: "omni"
   MODELS: >
     [
-      { "id": "zai-org/GLM-5.1", "description": "Upgraded 754B MoE for agentic coding, extended reasoning, and tool use.", "parameters": { "max_tokens": 32768 } },
-      { "id": "zai-org/GLM-5.1-FP8", "description": "FP8 GLM-5.1 for efficient agentic coding and reasoning inference.", "parameters": { "max_tokens": 32768 } },
+      { "id": "zai-org/GLM-5.1", "description": "Upgraded 754B MoE for agentic coding, extended reasoning, and tool use.", "parameters": { "max_tokens": 49152 } },
+      { "id": "zai-org/GLM-5.1-FP8", "description": "FP8 GLM-5.1 for efficient agentic coding and reasoning inference.", "parameters": { "max_tokens": 49152 } },
       { "id": "google/gemma-4-31B-it", "description": "Dense multimodal Gemma with 256K context, reasoning, and function calling." },
       { "id": "google/gemma-4-26B-A4B-it", "description": "Efficient multimodal MoE Gemma with 4B active params and 256K context." },
       { "id": "Qwen/Qwen3.5-9B", "description": "Dense multimodal hybrid with 262K context excelling at reasoning on-device." },

chart/env/prod.yaml

@@ -89,8 +89,8 @@ envVars:
   PUBLIC_LLM_ROUTER_ALIAS_ID: "omni"
   MODELS: >
     [
-      { "id": "zai-org/GLM-5.1", "description": "Upgraded 754B MoE for agentic coding, extended reasoning, and tool use.", "parameters": { "max_tokens": 32768 } },
-      { "id": "zai-org/GLM-5.1-FP8", "description": "FP8 GLM-5.1 for efficient agentic coding and reasoning inference.", "parameters": { "max_tokens": 32768 } },
+      { "id": "zai-org/GLM-5.1", "description": "Upgraded 754B MoE for agentic coding, extended reasoning, and tool use.", "parameters": { "max_tokens": 49152 } },
+      { "id": "zai-org/GLM-5.1-FP8", "description": "FP8 GLM-5.1 for efficient agentic coding and reasoning inference.", "parameters": { "max_tokens": 49152 } },
       { "id": "google/gemma-4-31B-it", "description": "Dense multimodal Gemma with 256K context, reasoning, and function calling." },
       { "id": "google/gemma-4-26B-A4B-it", "description": "Efficient multimodal MoE Gemma with 4B active params and 256K context." },
       { "id": "Qwen/Qwen3.5-9B", "description": "Dense multimodal hybrid with 262K context excelling at reasoning on-device." },