Update Ollama model recommendations

[RomneyDa]

Summary

• Added to Ollama: Qwen 3, Qwen 3 Coder, DeepSeek R1 (distilled), Gemma 3, Phi 4
• Removed from Ollama: CodeLlama, WizardCoder, Phind CodeLlama, old DeepSeek Coder (v1), Llama 3 (pre-3.1), Granite Code
• Updated onboarding default chat model: Llama 3.1 8B → Qwen 3 8B
• Updated tool support detection for gemma3, phi4
• Updated docs and provider descriptions with modern model examples

Removed models are still available via other providers (LM Studio, llama.cpp, etc).

Test plan

• Verify "Add Model" form shows new models when Ollama is selected
• Verify onboarding local tab uses qwen3:8b for chat
• Verify tool use detection works for gemma3 and phi4 models

---

Summary by cubic

Refreshes Ollama model recommendations and sets Qwen 3 8B as the default local chat model. Updates the Add Model UI, provider help text, and docs; adds tool-use detection for gemma3 and phi4.

• New Features

  • Add Model UI and defaults: adds Qwen 3, Qwen 3 Coder, DeepSeek R1 (distilled), Gemma 3, and Phi 4; sets qwen3:8b as the onboarding chat model; removes CodeLlama, WizardCoder, Phind CodeLlama, DeepSeek Coder v1, pre-3.1 Llama 3, and Granite Code from Ollama.
  • Docs and tooling: examples and provider tips now use qwen3:8b; VS Code schema copy and llm-info open-source presets updated; tool-use detection recognizes gemma3 and phi4.

• Migration

  • No breaking changes; existing configs continue to work.
  • Removed models remain available via lmstudio or llama.cpp. To use the new default, set your chat model to qwen3:8b or rerun onboarding.

^{Written for commit 385aacc. Summary will update on new commits.}

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
continue-docs	🟢 Ready	View Preview	Mar 27, 2026, 11:02 PM

[cubic-dev-ai]

No issues found across 7 files

[RomneyDa]

I'd double check qwen3 actually works decently w agent mode before merging this