Bug

The image tool rejects Ollama cloud vision models (e.g. ollama/qwen3-vl:235b-instruct-cloud, ollama/gemini-3-flash-preview) with Unknown model errors, even when these models are registered in the Ollama provider config with "input": ["text", "image"].

Steps to Reproduce

1. Configure an Ollama cloud vision model in models.providers.ollama.models with "input": ["text", "image"]
2. Set agents.defaults.imageModel to the Ollama cloud model (e.g. ollama/qwen3-vl:235b-instruct-cloud)
3. Call the image tool with any image
4. Get: Unknown model: ollama/qwen3-vl:235b-instruct-cloud

Expected

The image tool should resolve models from the provider config registry, not just a hardcoded allowlist. If a model is declared with image input support, it should be usable.

Workaround

Using ollama run via CLI works fine — the models support vision, it's just the tool's model resolution that blocks them.

Environment

• OpenClaw: 2026.4.12+
• Ollama provider with cloud models
• Config: models.providers.ollama.models[].input includes "image"