Prove the lift · AI model performance evaluation

llama-3.1-8b vs gpt-4o-mini

-2-point lift for llama-3.1-8b on insurance-extraction-eval

Test summary

This benchmark evaluates llama-3.1-8b against gpt-4o-mini on the insurance-extraction-eval v3 dataset (5 rows), scored by an LLM-as-judge (gpt-4o-mini). llama-3.1-8b reached a 52% pass rate versus 54% for gpt-4o-mini — a -2-point lift.

Base model
gpt-4o-mini
via openai · 54% pass rate
Refined model
llama-3.1-8b
via ollama · 52% pass rate · 1.4s p95
Base pass rate
54%
Refined pass rate
52%
Refined p95
1.4s
Est. tokens saved
688
0pts
no lift — the base model held its ground
0%
Base model
0%
Invoked-refined
3 improved0 regressed2 unchanged

Per-row results

1Coinsurance 80% clause recovery on an under-insured building0%100%+100
2Pro-rata distribution across layered carriers on a covered loss100%100%0
3Named-peril coverage determination under a basic HO form100%100%0
4Suit-limitation clause: is a late-filed claim contractually barred?0%100%+100
5Aggregate deductible across a loss sequence under a per-occurrence + annual cap0%100%+100
Prove the lift on your own data.

Run any two models over your tasks, judged automatically, and see the improvement — in seconds.

Try Invoked
Generated with Invoked · Browse all benchmarks