Prove the lift · AI model performance evaluation

gemma-3-27b vs gpt-oss-20b

+2-point lift for gemma-3-27b on ecommerce-catalog-norm

Test summary

This benchmark evaluates gemma-3-27b against gpt-oss-20b on the ecommerce-catalog-norm v3 dataset (5 rows), scored by an LLM-as-judge (claude-haiku-4-5, run locally). gemma-3-27b reached a 49% pass rate versus 47% for gpt-oss-20b — a +2-point lift.

Base model
gpt-oss-20b
via openai · 47% pass rate
Refined model
gemma-3-27b
via google · 49% pass rate · 3.3s p95
Base pass rate
47%
Refined pass rate
49%
Refined p95
3.3s
Est. tokens saved
1.3K
+0pts
refined beats base on 5 rows
0%
Base model
0%
Invoked-refined
0 improved2 regressed2 unchanged

Per-row results

1Classify a listing into the right category node100%0%-100
2Extract material + color from a product title100%0%-100
3Deduplicate two near-identical SKUs0%0%0
4Normalize a free-text size into a standard enum0%0%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