Prove the lift · AI model performance evaluation

llama-4-maverick vs o4-mini

-6-point lift for llama-4-maverick on code-review-bench

Test summary

This benchmark evaluates llama-4-maverick against o4-mini on the code-review-bench v3 dataset (5 rows), scored by an LLM-as-judge (claude-haiku-4-5). llama-4-maverick reached a 29% pass rate versus 35% for o4-mini — a -6-point lift.

Base model
o4-mini
via openai · 35% pass rate
Refined model
llama-4-maverick
via meta · 29% pass rate · 1.7s p95
Base pass rate
35%
Refined pass rate
29%
Refined p95
1.7s
Est. tokens saved
—
−0pts
no lift — the base model held its ground
0%
Base model
0%
Invoked-refined
▲3 improved▼1 regressed●1 unchanged

Per-row results

1Note a broken null-check after a refactor0% → 100%+100
2Spot a race in a read-modify-write on shared state0% → 100%+100
3Catch an unsanitized input reaching a SQL string100% → 0%-100
4Identify an N+1 query in an ORM call site0% → 0%0
5Detect a missing await on an async DB write0% → 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