Prove the lift · AI model performance evaluation

mistral-large vs gpt-4-turbo

+22-point lift for mistral-large on contract-clause-qa

Test summary

This benchmark evaluates mistral-large against gpt-4-turbo on the contract-clause-qa v2 dataset (5 rows), scored by an LLM-as-judge (llama-3.1-8b). mistral-large reached a 54% pass rate versus 32% for gpt-4-turbo — a +22-point lift.

Base model
gpt-4-turbo
via openai · 32% pass rate
Refined model
mistral-large
via mistral · 54% pass rate · 2.5s p95
Base pass rate
32%
Refined pass rate
54%
Refined p95
2.5s
Est. tokens saved
3.8K
+0pts
refined beats base on 5 rows
0%
Base model
0%
Invoked-refined
1 improved0 regressed4 unchanged

Per-row results

1Locate the termination-for-convenience terms100%100%0
2Find the liability cap and its exceptions0%0%0
3Determine indemnification scope and carve-outs100%100%0
4Extract the auto-renewal notice window in days0%0%0
5Identify the governing-law clause and jurisdiction0%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