Platform & company
AI models and providers
Which AI models power 8rney, how provider routing works, and what each thinking tier means for your research
8rney uses multiple frontier AI models under the hood, each tuned for different points on the speed-versus-depth spectrum. This page explains which model to choose and what each thinking tier means for your research.
Models#
GPT-5.5#
The default model for all new workspaces. Delivers fast, high-quality answers with configurable reasoning depth. Follow-up questions continue from the same reasoning context, so you don't restart from scratch when refining a question.
When to use: Day-to-day research across all practice areas. Best when your team needs quick turnaround and consistent latency.
Claude Sonnet 4.6#
A distinct model architecture with adaptive extended thinking that the model controls internally. Thinking summaries are surfaced in the activity feed so you can confirm that deeper reasoning occurred without wading through raw chain-of-thought.
When to use: When you want a second opinion from a different model, or when Claude's legal-reasoning style produces better results for your jurisdiction.
Claude Opus 4.6#
The most capable model on the platform. Adds a Max thinking tier on top of the standard Expert tier — the deepest reasoning available.
When to use: Your most demanding research tasks — contested litigation, multi-jurisdiction complex deals, constitutional questions, or any matter where a missed authority carries material risk.
Thinking tiers#
All providers map onto a five-tier pipeline mode system. Not every tier is available on every provider.
| Mode | Label shown in UI | GPT-5.5 | Claude Sonnet 4.6 | Claude Opus 4.6 |
|---|---|---|---|---|
fast | Fast | Low reasoning | Low adaptive | Low adaptive |
balanced | Balanced | Medium reasoning | Medium adaptive | Medium adaptive |
deep | Deep | High reasoning | High adaptive + summaries | High adaptive |
expert | Expert | Maximum reasoning | Max adaptive | Extra-high adaptive |
max | Max | — | — | Deepest reasoning (Opus only) |
Max mode is only shown in the composer when the Claude Opus provider is selected. It is hidden for GPT-5.5 and Claude Sonnet because those providers do not expose a higher tier beyond Expert.
How provider routing works#
When you select a model in the research settings, 8rney maps your chosen thinking tier to that model's reasoning capability and routes the request accordingly. If you switch models while on a tier that the new model does not support (for example, Max on Sonnet), the platform automatically steps down to Expert so your research continues without interruption.
Model selection is saved per-user — the platform remembers your last choice across sessions.
On this page