ProviderModelNotFoundError suggests the identical model string; multi-segment model keys fail as opaque "Unexpected server error"

Description

When a custom OpenAI-compatible provider defines model keys that themselves contain a slash (e.g. NVIDIA NIM: nvidia/nemotron-3.5-lightning-30b-a3b), opencode run -m <model-key> without the provider prefix fails in two unhelpful ways:

1. Without --print-logs, the CLI prints only Unexpected server error — nothing in the normal logs points at the cause (the log line ends at llm runtime selected).
2. With --print-logs, the underlying error is self-contradictory:

ProviderModelNotFoundError: Model not found: nvidia/nemotron-3.5-lightning-30b-a3b. Did you mean: nvidia/nemotron-3.5-lightning-30b-a3b?

The "not found" model and the suggestion are the identical string, so the message offers no actionable hint. The actual fix is to pass the provider namespace explicitly (nim/nvidia/nemotron-3.5-lightning-30b-a3b), which the error never suggests.

Expected

• -m <multi-segment-model-key> should either resolve, or fail with a clear message like "model key is ambiguous — prefix the provider: <provider>/<key>".
• The "Did you mean" suggestion should not repeat the exact unmatched string.
• A model-resolution failure should not surface to the user as a generic Unexpected server error.

Steps to reproduce

1. ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "nim": {
      "npm": "@ai-sdk/openai-compatible",
      "options": { "baseURL": "https://integrate.api.nvidia.com/v1", "apiKey": "{env:EXAMPLE_KEY}" },
      "models": { "nvidia/nemotron-3.5-lightning-30b-a3b": { "name": "Nemotron 3.5 Lightning 30B" } }
    }
  }
}

2. opencode run -m nvidia/nemotron-3.5-lightning-30b-a3b "Reply OK" → Unexpected server error
3. opencode run --print-logs -m nvidia/nemotron-3.5-lightning-30b-a3b "Reply OK" → ProviderModelNotFoundError: Model not found: nvidia/nemotron-3.5-lightning-30b-a3b. Did you mean: nvidia/nemotron-3.5-lightning-30b-a3b?
4. opencode run -m nim/nvidia/nemotron-3.5-lightning-30b-a3b "Reply OK" → works

Environment

• OpenCode 1.18.30 (opencode-ai npm package), Windows 11 (build 10.0.26200)
• Provider npm: @ai-sdk/openai-compatible; provider id: nim
• Possibly related: #39926 (parseModel splitting), #47690 (double-prefixing of model IDs)

Plugins: none in use.