Prove the lift · AI model performance evaluation

claude-sonnet-5 vs claude-haiku-4-5

+31-point lift for claude-sonnet-5 on financial-doc-extract

Test summary

This benchmark evaluates claude-sonnet-5 against claude-haiku-4-5 on the financial-doc-extract v2 dataset (5 rows), scored by an LLM-as-judge (llama-3.1-8b). claude-sonnet-5 reached a 82% pass rate versus 51% for claude-haiku-4-5 — a +31-point lift.

Base model
claude-haiku-4-5
via anthropic · 51% pass rate
Refined model
claude-sonnet-5
via anthropic · 82% pass rate · 1.4s p95
Base pass rate
51%
Refined pass rate
82%
Refined p95
1.4s
Est. tokens saved
6.9K
+0pts
refined beats base on 5 rows
0%
Base model
0%
Invoked-refined
▲4 improved▼0 regressed●1 unchanged

Per-row results

1Extract net revenue and YoY change from an earnings PDF0% → 100%+100
2Determine the reporting period from a 10-Q header0% → 100%+100
3Pull the diluted EPS from an income statement100% → 100%0
4Extract total debt from a balance sheet0% → 100%+100
5Identify a going-concern note in the footnotes0% → 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