LLM Rankings
Six models we track closely. Raw = the model alone. Juiced = same model through the DocketRouter legal model.
Real, blinded Texas appellate cases decided after every model's training cutoff. Score = weighted issue / standard / authority / application / outcome / procedure, with any fabricated citation capping the item at 25%. Methodology →
| # | Model | Provider | Raw | Juiced | Δ | Items | Fabricated-cite items (raw / juiced) |
|---|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.5 anthropic/claude-sonnet-4.5 | Anthropic | 39% | - | - | 34 | 17 |