lintl runs on Amazon Bedrock (and optional USA Mantle models). You pick a model for each chat and, when the model supports it, an **Intelligence** level (Instant / Medium / High) that controls how hard it thinks — not which brand you picked.
Choosing a model
- On desktop, open the model control in the composer.
- On mobile, open **+** → **Model**.
- Defaults: **Claude Sonnet 4.6** for everyday UK work. **Claude Sonnet 5.5** and **Claude Opus 5.5** are also in the EU list.
Prefer a lighter model for drafting and tidy-ups. Reserve Opus-class or High intelligence when the planning judgement actually needs it.
Models you may see
**EU/UK (default family):** Claude Opus / Sonnet / Haiku, Amazon Nova Pro and Nova 2 Lite, GPT OSS, Voxtral, Kimi, Qwen (including vision variants).
**USA-hosted** (only when your account enables them and you accept the relevant consent): Mantle GPT-5.x, DeepSeek R1, Nova Omni, Claude Fable 5. Some GPT-5.x ids may show as **Coming soon** until AWS entitlement is ready.
Intelligence (Instant / Medium / High)
When the selected model supports effort levels:
- **Instant** — fast replies; good for drafting
- **Medium** — balanced default
- **High** — slower, deeper reasoning; use when the planning call matters
Settings → Models
Account → Settings → **Models** lets you hide models you never use. Opting out of a model does not delete chats that already used it.