Prove the lift · AI model performance evaluation

mixtral-8x7b vs gpt-4o-mini

-7-point lift for mixtral-8x7b on code-review-bench

Test summary

This benchmark evaluates mixtral-8x7b against gpt-4o-mini on the code-review-bench v1 dataset (6 rows), scored by an LLM-as-judge (gpt-4o-mini). mixtral-8x7b reached a 29% pass rate versus 36% for gpt-4o-mini — a -7-point lift.

Base model
gpt-4o-mini
via openai · 36% pass rate
Refined model
mixtral-8x7b
via mistral · 29% pass rate · 1.3s p95
Base pass rate
36%
Refined pass rate
29%
Refined p95
1.3s
Est. tokens saved
517
−0pts
no lift — the base model held its ground
0%
Base model
0%
Invoked-refined
▲0 improved▼1 regressed●5 unchanged

Per-row results

1Identify an N+1 query in an ORM call site100% → 100%0
2Spot a race in a read-modify-write on shared state0% → 0%0
3Find a resource leak on an early-return path100% → 0%-100
4Catch an unsanitized input reaching a SQL string0% → 0%0
5Detect a missing await on an async DB write0% → 0%0
6Flag an off-by-one in a pagination loop100% → 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