Prove the lift · AI model performance evaluation

gpt-4o vs claude-haiku-4-5

+16-point lift for gpt-4o on financial-doc-extract

Test summary

This benchmark evaluates gpt-4o against claude-haiku-4-5 on the financial-doc-extract v3 dataset (5 rows), scored by an LLM-as-judge (llama-3.1-8b). gpt-4o reached a 44% pass rate versus 28% for claude-haiku-4-5 — a +16-point lift.

Base model
claude-haiku-4-5
via anthropic · 28% pass rate
Refined model
gpt-4o
via openai · 44% pass rate · 3.8s p95
Base pass rate
28%
Refined pass rate
44%
Refined p95
3.8s
Est. tokens saved
4.2K
+0pts
refined beats base on 5 rows
0%
Base model
0%
Invoked-refined
▲2 improved▼0 regressed●3 unchanged

Per-row results

1Determine the reporting period from a 10-Q header0% → 100%+100
2Pull the diluted EPS from an income statement0% → 0%0
3Identify a going-concern note in the footnotes0% → 0%0
4Extract total debt from a balance sheet0% → 100%+100
5Extract net revenue and YoY change from an earnings PDF0% → 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 Invoked
Generated with Invoked · Browse all benchmarks