Prove the lift · AI model performance evaluation
claude-opus-4-8 vs claude-opus-4-6
+19-point lift for claude-opus-4-8 on legal-research-qa
Test summary
This benchmark evaluates claude-opus-4-8 against claude-opus-4-6 on the legal-research-qa v3 dataset (5 rows), scored by an LLM-as-judge (llama-3.1-8b). claude-opus-4-8 reached a 75% pass rate versus 56% for claude-opus-4-6 — a +19-point lift.
- Base model
- claude-opus-4-6
- via anthropic · 56% pass rate
- Refined model
- claude-opus-4-8
- via anthropic · 75% pass rate · 3.2s p95
Base pass rate
56%
Refined pass rate
75%
Refined p95
3.2s
Est. tokens saved
8.1K
+0pts
refined beats base on 5 rows
0%
Base model0%
Invoked-refined▲2 improved▼1 regressed●1 unchanged
Per-row results
1Extract the standard of review applied0% → 100%+100
2Identify the controlling statute in a fact pattern100% → 100%0
3Determine whether a precedent was overruled0% → 100%+100
4Summarize the holding of a cited appellate case100% → 0%-100
Prove the lift on your own data.
Run any two models over your tasks, judged automatically, and see the improvement — in seconds.
Try InvokedGenerated with Invoked · Browse all benchmarks