Added Vision Model - OpenCode is extremely buggy - text models are operating fine

[surapuramakhil]

Description

Summary

Add two models

1. GLM - 5 Text only model
2. GLM - 4.6v

via customer provider Opencode.jsonc

....

"models": {
        "Qwen/Qwen3-VL-235B-A22B-Instruct": {
          "name": "Qwen: Qwen3 VL 235B Instruct",
          "attachment": true,
          "modalities": {
            "input": [
              "text",
              "image"
            ],
            "output": [
              "text"
            ]
          }
        },
        "zai-org/GLM-5": {
          "name": "z-ai glm 5"
        },
        "zai-org/GLM-4.6v": {
          "name": "z-ai glm 4.6v",
          "attachment": true,
          "modalities": {
            "input": [
              "text",
              "image"
            ],
            "output": [
              "text"
            ]
          }
        }
      }

...

GLM 5 works great buth both Qwen & GLM 4.6v struggle getting responses , see lag of 3 -7 seconds sometimes more than that.

Steps to reproduce

1. add provided models - in openCode.jsonc
2. restart Opencode
3. those model start appearing - select one of them and try having a chat
4. you find extreme lag with those models

I thought it might be inference issue but its not, inferences is as fast as other models

Expected behavior

it should be working as smooths as text models

Actual behavior

Laggy 8 - 10 seconds processing time for each step

Plugins

No response

OpenCode version

No response

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response