Prove the lift · AI model performance evaluation
claude-sonnet-4-5 vs minimax-m3
+7-point lift for claude-sonnet-4-5 on medical-coding-extract
Test summary
This benchmark evaluates claude-sonnet-4-5 against minimax-m3 on the medical-coding-extract v2 dataset (5 rows), scored by an LLM-as-judge (gpt-4o-mini). claude-sonnet-4-5 reached a 31% pass rate versus 24% for minimax-m3 — a +7-point lift.
- Base model
- minimax-m3
- via minimax · 24% pass rate
- Refined model
- claude-sonnet-4-5
- via anthropic · 31% pass rate · 2.4s p95
Base pass rate
24%
Refined pass rate
31%
Refined p95
2.4s
Est. tokens saved
6.7K
+0pts
refined beats base on 5 rows
0%
Base model0%
Invoked-refined▲1 improved▼0 regressed●4 unchanged
Per-row results
1Determine the correct E/M level from a visit note0% → 0%0
2Identify a contraindication from a problem list0% → 100%+100
3Extract medication + dosage from a clinical note100% → 100%0
4Map a discharge summary to the primary ICD-10 code100% → 100%0
5Pull the procedure code from an operative report0% → 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 InvokedGenerated with Invoked · Browse all benchmarks