Prove the lift · AI model performance evaluation
claude-opus-4-6 vs claude-opus-4-8
+18-point lift for claude-opus-4-6 on legal-research-qa
Test summary
This benchmark evaluates claude-opus-4-6 against claude-opus-4-8 on the legal-research-qa v1 dataset (5 rows), scored by an LLM-as-judge (gpt-4o-mini). claude-opus-4-6 reached a 53% pass rate versus 35% for claude-opus-4-8 — a +18-point lift.
- Base model
- claude-opus-4-8
- via anthropic · 35% pass rate
- Refined model
- claude-opus-4-6
- via anthropic · 53% pass rate · 0.5s p95
Base pass rate
35%
Refined pass rate
53%
Refined p95
0.5s
Est. tokens saved
5.2K
+0pts
refined beats base on 5 rows
0%
Base model0%
Invoked-refined▲3 improved▼0 regressed●1 unchanged
Per-row results
1Summarize the holding of a cited appellate case0% → 100%+100
2Identify the controlling statute in a fact pattern0% → 100%+100
3Determine whether a precedent was overruled0% → 100%+100
4Extract the standard of review applied0% → 0%0
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