Prove the lift · AI model performance evaluation

claude-opus-4-6 vs claude-opus-4-8

-1-point lift for claude-opus-4-6 on financial-doc-extract

Test summary

This benchmark evaluates claude-opus-4-6 against claude-opus-4-8 on the financial-doc-extract v3 dataset (5 rows), scored by an LLM-as-judge (claude-haiku-4-5). claude-opus-4-6 reached a 42% pass rate versus 43% for claude-opus-4-8 — a -1-point lift.

Base model
claude-opus-4-8
via anthropic · 43% pass rate
Refined model
claude-opus-4-6
via anthropic · 42% pass rate · 3.1s p95
Base pass rate
43%
Refined pass rate
42%
Refined p95
3.1s
Est. tokens saved
539
0pts
no lift — the base model held its ground
0%
Base model
0%
Invoked-refined
0 improved1 regressed4 unchanged

Per-row results

1Identify a going-concern note in the footnotes100%100%0
2Extract net revenue and YoY change from an earnings PDF0%0%0
3Extract total debt from a balance sheet100%0%-100
4Pull the diluted EPS from an income statement100%100%0
5Determine the reporting period from a 10-Q header100%100%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