Prove the lift · AI model performance evaluation
llama-3.1-405b vs grok-4
+16-point lift for llama-3.1-405b on financial-doc-extract
Test summary
This benchmark evaluates llama-3.1-405b against grok-4 on the financial-doc-extract v1 dataset (5 rows), scored by an LLM-as-judge (llama-3.1-8b). llama-3.1-405b reached a 69% pass rate versus 53% for grok-4 — a +16-point lift.
- Base model
- grok-4
- via xai · 53% pass rate
- Refined model
- llama-3.1-405b
- via meta · 69% pass rate · 0.9s p95
Base pass rate
53%
Refined pass rate
69%
Refined p95
0.9s
Est. tokens saved
10.7K
+0pts
refined beats base on 5 rows
0%
Base model0%
Invoked-refined▲5 improved▼0 regressed●0 unchanged
Per-row results
1Extract total debt from a balance sheet0% → 100%+100
2Extract net revenue and YoY change from an earnings PDF0% → 100%+100
3Determine the reporting period from a 10-Q header0% → 100%+100
4Identify a going-concern note in the footnotes0% → 100%+100
5Pull the diluted EPS from an income statement0% → 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