Prove the lift · AI model performance evaluation
grok-4 vs claude-3-7-sonnet
+19-point lift for grok-4 on financial-doc-extract
Test summary
This benchmark evaluates grok-4 against claude-3-7-sonnet on the financial-doc-extract v2 dataset (5 rows), scored by an LLM-as-judge (claude-haiku-4-5). grok-4 reached a 63% pass rate versus 44% for claude-3-7-sonnet — a +19-point lift.
- Base model
- claude-3-7-sonnet
- via anthropic · 44% pass rate
- Refined model
- grok-4
- via xai · 63% pass rate · 2.5s p95
Base pass rate
44%
Refined pass rate
63%
Refined p95
2.5s
Est. tokens saved
10.5K
+0pts
refined beats base on 5 rows
0%
Base model0%
Invoked-refined▲4 improved▼1 regressed●0 unchanged
Per-row results
1Identify a going-concern note in the footnotes0% → 100%+100
2Pull the diluted EPS from an income statement0% → 100%+100
3Determine the reporting period from a 10-Q header0% → 100%+100
4Extract net revenue and YoY change from an earnings PDF100% → 0%-100
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 InvokedGenerated with Invoked · Browse all benchmarks