/models command shows preset list instead of actual LM Studio local models #25351

Description

Describe the bug
When using LM Studio as local provider, the /models command displays hardcoded preset models from models.dev (like "GPT OSS 20B", "Qwen3 30B A3B 2507") instead of querying the local LM Studio API endpoint http://127.0.0.1:1234/v1/models to get actually loaded models.

To Reproduce

1. Start LM Studio and load a model (e.g., gemma-4-26b-a4b-it)
2. Enable Local Server (default port 1234)
3. Configure opencode.json with lmstudio provider
4. Run /models in OpenCode Desktop/TUI

Expected behavior
/models should query http://127.0.0.1:1234/v1/models and display only models actually available in local LM Studio instance.

Actual behavior
Shows preset models from models.dev that user does not have downloaded.

Evidence
Direct API call returns correct local models:

curl http://127.0.0.1:1234/v1/models

About the duplicate flag: #18011 reports that LM Studio shows only 3/9 models (partial list). This issue reports that /models shows a completely different preset list (from models.dev) rather than querying the API at all. These are different symptoms: #18011 gets some models from the API but not all; this issue gets zero models from the API. #2047 and #23327 are related (LM Studio model refresh issues) but do not address the preset vs API discrepancy.