Prove the lift · AI model performance evaluation

llama-3.1-405b vs claude-3-5-haiku

+21-point lift for llama-3.1-405b on support-ticket-triage

Test summary

This benchmark evaluates llama-3.1-405b against claude-3-5-haiku on the support-ticket-triage v3 dataset (5 rows), scored by an LLM-as-judge (claude-haiku-4-5). llama-3.1-405b reached a 47% pass rate versus 26% for claude-3-5-haiku — a +21-point lift.

Base model
claude-3-5-haiku
via anthropic · 26% pass rate
Refined model
llama-3.1-405b
via meta · 47% pass rate · 3.1s p95
Base pass rate
26%
Refined pass rate
47%
Refined p95
3.1s
Est. tokens saved
13.1K
+0pts
refined beats base on 5 rows
0%
Base model
0%
Invoked-refined
2 improved0 regressed3 unchanged

Per-row results

1Detect churn-risk sentiment in a renewal thread0%0%0
2Classify: billing dispute vs refund request vs cancellation100%100%0
3Priority-tag an outage report vs a feature question0%100%+100
4Summarize a 12-message thread into a one-line disposition0%100%+100
5Route a multi-part ticket to the correct queue100%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