Medicine
MedHELM MedCalc-Bench
MedHELM medcalc_bench: per-(model, item) medcalc_bench_accuracy in {0,1} (1=correct) on 1000 clinical calculations from a patient note (risk scores, dosing, lab-derived values). 13 models from the MedHELM v4.0.0 release.
992items
13subjects
100%observed
Apache-2.0license
medicinedomain
textmodality
Response matrix
Fit to width. Hover for subject & item; click a cell for details.
Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect
Subjects
- 1deepseek-ai/deepseek-r10.348
- 2google/gemini-2.5-pro-preview-05-060.348
- 3openai/gpt-5-mini-2025-08-070.341
- 4openai/o3-mini-2025-01-310.34
- 5openai/o4-mini-2025-04-160.34
- 6openai/gpt-5-2025-08-070.333
- 7anthropic/claude-3-5-sonnet-202410220.218
- 8anthropic/claude-3-7-sonnet-202502190.21
- 9openai/gpt-4o-2024-05-130.188
- 10google/gemini-2.0-flash-0010.158
- 11openai/gpt-4o-mini-2024-07-180.154
- 12google/gemini-1.5-pro-0010.12
- 13meta/llama-3.3-70b-instruct0.113