Methodology
This site is an index, not a single evaluation. It gathers the current public results of every healthcare benchmark that frontier models are still being tested on, and its discipline is provenance: each score appears exactly as its source published it, with that source named on the row. Where the index's maintainers run the underlying leaderboard themselves, the entry says so.
Where the numbers come from
| Sister boards | HealthBench Professional, HealthBench Hard, and Health Optimization Bench results come from the dedicated leaderboard sites this index's maintainers also run, at healthbenchprofessional.com, healthbenchhard.ai, and healthoptimizationbench.com, where each protocol is documented. |
|---|---|
| Publisher sources | Scores published by a benchmark's own maintainer or by a model vendor: system cards, deployment safety pages, official leaderboards. Recorded as the source shows them, marked vendor-reported where the vendor graded its own model. |
| Third-party trackers | Independent leaderboard sites that either re-run models or mirror published numbers. The tracker is named on each entry, and mirroring trackers inherit the vendor-reported label. |
| Inclusion rule | A benchmark earns a page only while at least one current frontier model has a public score on it. Boards with no frontier activity are listed as retired rather than carried with stale numbers. |
What can be compared
Within one benchmark's table, rows share a task set but not always a configuration: when a source mixes grader versions or settings, the config column says so, and rows with different configs should be read loosely. Across benchmarks, nothing is comparable at all. A 0.6 on a rubric benchmark, a 46 percent on a hard subset, and an accuracy figure on an exam-style test are three different measurements that happen to share a page, which is why this site never averages them into a single healthcare number.
Limits
An index inherits the flaws of its sources. Vendor-reported scores favor the vendor's configuration, model-graded rubrics carry grader bias, and some sources update irregularly, so a date on a row is part of the data, not decoration. When a source and this index disagree, the source is authoritative, and a correction is a dated entry on the updates page.
Refresh cadence
The index is refreshed when a frontier model ships, when a tracked source publishes new results, or when a new benchmark qualifies. The current snapshot, covering 18 benchmarks, is from August 16, 2026; the full history is on the data page.