# HealthBench

License: permissive (MIT); Mdexa hosting: link-only. first-hand data-scope statement(s) allow commercial use and redistribution

- Commercial use: Yes · Derivatives: Yes · Redistribution: Yes · File on Mdexa: No
- License status: Permissive (MIT)
- Mdexa hosting: Link only — License permissive: first-hand data-scope statement(s) allow commercial use and redistribution. Link only.
- Last verified: 2026-11-14
- Card: https://mdexa.com/datasets/healthbench/

## Key facts

- Domain: Clinical medicine
- Content type: Question banks and benchmarks
- Size: —
- Language: English
- Modality: Text
- Publisher: openai
- Year: —

## Commercial use of HealthBench

- Commercial use: Yes.
- License status: Permissive.
- Conditions: Notice kept
Verdicts record what sources state and are not legal advice.

## HealthBench license evidence

first-hand data-scope statement(s) allow commercial use and redistribution

- **Hugging Face card metadata** — MIT: Statement found in snapshot s09 (readme) — "license: mit" — https://huggingface.co/datasets/openai/healthbench (2026-11-12; Page opened)

## Source and lineage

—

## Known issues with HealthBench

—

## Preflight

Not run (data not ingested).

## Downloading HealthBench and API access

Mdexa does not serve files. Original source:
- https://huggingface.co/datasets/openai/healthbench

### Access recipe

- .gitattributes · 3 KB · https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/.gitattributes
- 2025-05-07-06-14-12_oss_eval.jsonl · 60.3 MB · sha256 e99dd3c6372c10d6fcc5e385c5fae69d0dd40392dae56836ef9493ae324ecd2f · https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/2025-05-07-06-14-12_oss_eval.jsonl
- 2025-05-07-06-14-12_oss_meta_eval.jsonl · 136.2 MB · sha256 6a343ed3fcb6f28353c2c6c157ae30ddf2db874c947b27ba474ad54da9c5dbac · https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/2025-05-07-06-14-12_oss_meta_eval.jsonl
- README.md · 1 KB · https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/README.md
- consensus_2025-05-09-20-00-46.jsonl · 36.9 MB · sha256 fabe37553327238928b05357d8b309737b5d426d6dee432a3376da618d0969d6 · https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/consensus_2025-05-09-20-00-46.jsonl
- hard_2025-05-08-21-00-10.jsonl · 12.6 MB · sha256 b0320430e5cd974e746585594c1dd10b5a3fc2aff9c72b26106c2c4a069d74e9 · https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/hard_2025-05-08-21-00-10.jsonl

```sh
curl -L --create-dirs -o ".gitattributes" "https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/.gitattributes"
curl -L --create-dirs -o "2025-05-07-06-14-12_oss_eval.jsonl" "https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/2025-05-07-06-14-12_oss_eval.jsonl"
echo "e99dd3c6372c10d6fcc5e385c5fae69d0dd40392dae56836ef9493ae324ecd2f  2025-05-07-06-14-12_oss_eval.jsonl" | sha256sum -c
curl -L --create-dirs -o "2025-05-07-06-14-12_oss_meta_eval.jsonl" "https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/2025-05-07-06-14-12_oss_meta_eval.jsonl"
echo "6a343ed3fcb6f28353c2c6c157ae30ddf2db874c947b27ba474ad54da9c5dbac  2025-05-07-06-14-12_oss_meta_eval.jsonl" | sha256sum -c
curl -L --create-dirs -o "README.md" "https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/README.md"
curl -L --create-dirs -o "consensus_2025-05-09-20-00-46.jsonl" "https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/consensus_2025-05-09-20-00-46.jsonl"
echo "fabe37553327238928b05357d8b309737b5d426d6dee432a3376da618d0969d6  consensus_2025-05-09-20-00-46.jsonl" | sha256sum -c
curl -L --create-dirs -o "hard_2025-05-08-21-00-10.jsonl" "https://huggingface.co/datasets/openai/healthbench/resolve/40ee1968852fc57f625934251ac22be47077a8fb/hard_2025-05-08-21-00-10.jsonl"
echo "b0320430e5cd974e746585594c1dd10b5a3fc2aff9c72b26106c2c4a069d74e9  hard_2025-05-08-21-00-10.jsonl" | sha256sum -c
```

## Citing HealthBench

No recommended citation text.

## Machine-readable

- JSON: https://mdexa.com/datasets/healthbench.json
- 2026-11-14 Added via the automated pipeline and passed all checks (gates, verbatim quotes, rule engine, two independent extractions).

## Continue in this database

This record is one of 109 processed and verified records (datasets, data resources and open models). To find others, call `https://mcp.mdexa.com/v1/datasets?q=<words>` or connect the MCP server at `https://mcp.mdexa.com/mcp`; no key is needed for 20 calls per minute. Rows are under 1 KB; an empty result means "not listed". For a list of items, one batch check (free account) returns a report. Guide: /agents.md. Similar records: CBLUE (https://mdexa.com/datasets/cblue/); CMB (Chinese Medical Benchmark) (https://mdexa.com/datasets/cmb/); CMExam (https://mdexa.com/datasets/cmexam/).