Prove the lift · AI model performance evaluation

mistral-large vs llama-3.1-70b

+1-point lift for mistral-large on support-ticket-triage

Test summary

This benchmark evaluates mistral-large against llama-3.1-70b on the support-ticket-triage v2 dataset (6 rows), scored by an LLM-as-judge (claude-haiku-4-5, run locally). mistral-large reached a 31% pass rate versus 30% for llama-3.1-70b — a +1-point lift.

Base model
llama-3.1-70b
via ollama · 30% pass rate
Refined model
mistral-large
via mistral · 31% pass rate · 2.6s p95
Base pass rate
30%
Refined pass rate
31%
Refined p95
2.6s
Est. tokens saved
4.8K
+0pts
refined beats base on 6 rows
0%
Base model
0%
Invoked-refined
3 improved1 regressed2 unchanged

Per-row results

1Extract the affected order id from a free-text complaint0%100%+100
2Detect churn-risk sentiment in a renewal thread100%100%0
3Summarize a 12-message thread into a one-line disposition0%0%0
4Priority-tag an outage report vs a feature question0%100%+100
5Route a multi-part ticket to the correct queue0%100%+100
6Classify: billing dispute vs refund request vs cancellation100%0%-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