Prove the lift · AI model performance evaluation

gpt-5.1 vs claude-opus-4-6

+20-point lift for gpt-5.1 on code-review-bench

Test summary

This benchmark evaluates gpt-5.1 against claude-opus-4-6 on the code-review-bench v1 dataset (7 rows), scored by an LLM-as-judge (llama-3.1-8b). gpt-5.1 reached a 58% pass rate versus 38% for claude-opus-4-6 — a +20-point lift.

Base model
claude-opus-4-6
via anthropic · 38% pass rate
Refined model
gpt-5.1
via openai · 58% pass rate · 2.7s p95
Base pass rate
38%
Refined pass rate
58%
Refined p95
2.7s
Est. tokens saved
6.8K
+0pts
refined beats base on 7 rows
0%
Base model
0%
Invoked-refined
▲4 improved▼1 regressed●2 unchanged

Per-row results

1Note a broken null-check after a refactor0% → 100%+100
2Flag an off-by-one in a pagination loop0% → 0%0
3Detect a missing await on an async DB write0% → 100%+100
4Find a resource leak on an early-return path0% → 100%+100
5Catch an unsanitized input reaching a SQL string100% → 0%-100
6Spot a race in a read-modify-write on shared state0% → 0%0
7Identify an N+1 query in an ORM call site0% → 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