Prove the lift · AI model performance evaluation

gemini-3-pro vs llama-4-maverick

-9-point lift for gemini-3-pro on financial-doc-extract

Test summary

This benchmark evaluates gemini-3-pro against llama-4-maverick on the financial-doc-extract v1 dataset (5 rows), scored by an LLM-as-judge (llama-3.1-8b). gemini-3-pro reached a 40% pass rate versus 49% for llama-4-maverick — a -9-point lift.

Base model
llama-4-maverick
via meta · 49% pass rate
Refined model
gemini-3-pro
via google · 40% pass rate · 2.9s p95
Base pass rate
49%
Refined pass rate
40%
Refined p95
2.9s
Est. tokens saved
1.4K
−0pts
no lift — the base model held its ground
0%
Base model
0%
Invoked-refined
▲1 improved▼0 regressed●4 unchanged

Per-row results

1Extract net revenue and YoY change from an earnings PDF100% → 100%0
2Determine the reporting period from a 10-Q header0% → 0%0
3Pull the diluted EPS from an income statement100% → 100%0
4Identify a going-concern note in the footnotes0% → 0%0
5Extract total debt from a balance sheet0% → 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