Prove the lift · AI model performance evaluation
grok-4 vs gemini-3-pro
+9-point lift for grok-4 on insurance-extraction-eval
Test summary
This benchmark evaluates grok-4 against gemini-3-pro on the insurance-extraction-eval v3 dataset (6 rows), scored by an LLM-as-judge (llama-3.1-8b, run locally). grok-4 reached a 61% pass rate versus 52% for gemini-3-pro — a +9-point lift.
- Base model
- gemini-3-pro
- via google · 52% pass rate
- Refined model
- grok-4
- via xai · 61% pass rate · 3.1s p95
Base pass rate
52%
Refined pass rate
61%
Refined p95
3.1s
Est. tokens saved
1.5K
+0pts
refined beats base on 6 rows
0%
Base model0%
Invoked-refined▲3 improved▼0 regressed●3 unchanged
Per-row results
1Named-peril coverage determination under a basic HO form0% → 100%+100
2Replacement cost vs ACV on a depreciated roof0% → 100%+100
3Pro-rata distribution across layered carriers on a covered loss0% → 100%+100
4Suit-limitation clause: is a late-filed claim contractually barred?100% → 100%0
5Coinsurance 80% clause recovery on an under-insured building100% → 100%0
6Aggregate deductible across a loss sequence under a per-occurrence + annual cap100% → 100%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