Prove the lift · AI model performance evaluation

deepseek-r1 vs llama-3.1-405b

+21-point lift for deepseek-r1 on medical-coding-extract

Test summary

This benchmark evaluates deepseek-r1 against llama-3.1-405b on the medical-coding-extract v2 dataset (5 rows), scored by an LLM-as-judge (llama-3.1-8b). deepseek-r1 reached a 76% pass rate versus 55% for llama-3.1-405b — a +21-point lift.

Base model
llama-3.1-405b
via meta · 55% pass rate
Refined model
deepseek-r1
via deepseek · 76% pass rate · 3.1s p95
Base pass rate
55%
Refined pass rate
76%
Refined p95
3.1s
Est. tokens saved
9.1K
+0pts
refined beats base on 5 rows
0%
Base model
0%
Invoked-refined
5 improved0 regressed0 unchanged

Per-row results

1Determine the correct E/M level from a visit note0%100%+100
2Pull the procedure code from an operative report0%100%+100
3Extract medication + dosage from a clinical note0%100%+100
4Map a discharge summary to the primary ICD-10 code0%100%+100
5Identify a contraindication from a problem list0%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