Prove the lift · AI model performance evaluation

gpt-4o-mini vs claude-sonnet-4-6

-17-point lift for gpt-4o-mini on contract-clause-qa

Test summary

This benchmark evaluates gpt-4o-mini against claude-sonnet-4-6 on the contract-clause-qa v1 dataset (6 rows), scored by an LLM-as-judge (claude-haiku-4-5). gpt-4o-mini reached a 50% pass rate versus 67% for claude-sonnet-4-6 — a -17-point lift.

Base model
claude-sonnet-4-6
via anthropic · 67% pass rate
Refined model
gpt-4o-mini
via openai · 50% pass rate · 3.1s p95
Base pass rate
67%
Refined pass rate
50%
Refined p95
3.1s
Est. tokens saved
—
−0pts
no lift — the base model held its ground
0%
Base model
0%
Invoked-refined
▲0 improved▼2 regressed●4 unchanged

Per-row results

1Identify the governing-law clause and jurisdiction100% → 0%-100
2Extract the auto-renewal notice window in days100% → 100%0
3Determine indemnification scope and carve-outs100% → 0%-100
4Find the liability cap and its exceptions0% → 0%0
5Locate the termination-for-convenience terms0% → 100%0
6Governing venue vs choice-of-law distinction100% → 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