Skip to main content
AIMSAI Measurement Science

Medicine

MedHELM MedMCQA

MedHELM med_mcqa: per-(model, item) exact_match in {0,1} (1=correct) on 1000 Indian medical-entrance (AIIMS/NEET-PG) multiple-choice questions. 13 models from the MedHELM v4.0.0 release.

1,000items
13subjects
100%observed
Apache-2.0license
medicinedomain
textmodality
Original sourcePaperBuild script← All benchmarks

Response matrix

Fit to width. Hover for subject & item; click a cell for details.

MedHELM MedMCQA response matrix: AI models (rows) against items (columns)
Correct (1)Incorrect (0)Unobserved

Scale: 1 = correct · 0 = incorrect

Subjects

  1. 1openai/gpt-5-2025-08-070.847
  2. 2google/gemini-2.5-pro-preview-05-060.834
  3. 3openai/o4-mini-2025-04-160.817
  4. 4openai/gpt-5-mini-2025-08-070.816
  5. 5openai/o3-mini-2025-01-310.766
  6. 6google/gemini-2.0-flash-0010.748
  7. 7deepseek-ai/deepseek-r10.745
  8. 8openai/gpt-4o-2024-05-130.744
  9. 9anthropic/claude-3-7-sonnet-202502190.74
  10. 10anthropic/claude-3-5-sonnet-202410220.74
  11. 11meta/llama-3.3-70b-instruct0.724
  12. 12google/gemini-1.5-pro-0010.71
  13. 13openai/gpt-4o-mini-2024-07-180.664
Full data on Hugging FaceBack to the gallery