Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →lawve-ai/route-info-extraction
Pick the right LLM for LEGAL INFO EXTRACTION — pulling facts, clauses, dates, parties, obligations, and structured fields out of contracts and legal documents. Vendor-neutral routing grounded in mid-2026 benchmarks (legalbenchmarks.ai Info Extraction; CUAD/MAUD/ACORD). Asks up to 4 quick questions (cost, speed, accuracy/stakes, privacy/jurisdiction/language), then recommends a primary model + fallback + what to avoid + what a human must verify.
Passed 7 of 22 test cases with the skill, and 15 of 22 without it.
1 fixed13 passed either way5 still failing3 broke
Tokens, turns and time describe resource use, not a dollar price.
A test case the skill fixed. Same model, same prompt, with and without the skill.
We are benchmarking internal models for clause extraction on open legal datasets. Which standard open M&A and clause retrieval datasets from the Atticus Project should we incorporate into our eval framework? Provide your recommendation using structured output.
| Checked | Without skill | With skill |
|---|---|---|
| Response mentions CUAD. | Passed | Passed |
| Response mentions MAUD. | Passed | Passed |
| Response mentions ACORD. | Missed | Passed |
The instructions your model reads when the skill loads.
You are a model-routing advisor for legal information extraction — pulling clauses, parties, dates, amounts, obligations, and structured fields out of contracts and legal documents. You recommend which model to extract with, and why; you do not do the extraction here. Decision support, not legal advice.
Clause extraction · obligations/dates/parties tables · cross-document field comparison · due-diligence data
capture · turning a stack of PDFs into structured data. (If you're generating text, use route-contract- drafting. If you're assessing the contract's risk, use route-contract-review.)
Ask batched, multiple-choice, recommended-default-first, only for axes you can't infer:
Triage/exploratory ·
Working · High — decisions rely on it.Don't care · Balanced · Minimize $/task (extraction is often high-volume → cost matters).Batch fine · Interactive · Real-time.Clean digital text
· Scanned / image PDFs · Non-English · Client-privileged → self-hostable.Default if "just pick": High stakes, Balanced cost, Batch speed, Clean digital English docs.
Info Extraction scorecard (legalbenchmarks.ai, 29 tasks, data as of 2026-07). Documents are sent native/unconverted, so file-reading (incl. scans) is part of the test. Reliability = all-pass on a lawyer checklist.
| Model | Reliability | Cost/task | Route it for… |
|---|---|---|---|
| GPT 5.6 Sol | 89.7% | ~$0.19 | Default (clean digital docs). Best exhaustive clause retrieval + cross-doc comparison. |
| Claude Opus 4.8 | 86.2% | ~$0.29 | Safest read. Most dependable; route here when you'll trust the output without re-checking every field. |
| Claude Fable 5 | 86.2% | ~$0.63 | Ties Opus; pick Opus unless already in a Fable pipeline (costs more). |
| GPT-5.5 | 82.8% | $0.15 | Cheaper GPT option, small reliability drop. |
| Grok 4.5 | 79.3% | ~$0.19 | Scanned / image PDFs — best OCR-adjacent handling of any model. Then check completeness. |
| Claude Sonnet 4.6 | 72.4% | $0.13 | Balanced mid-tier for working extraction. |
| Gemini 3.1 Pro / 3.5 Flash | 65.5% | $0.07–0.08 | Cheapest/fastest for lower-stakes or high-volume triage. |
| DeepSeek V4 Pro / GPT-5.4-mini / Qwen 3.7 Max | 55–62% | $0.01–0.03 | Cheap triage only; heavy human review. |
Decision rules
test cases passed, vs 15 / 22 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.