Open Dataset Specification

Standard AI Benchmark Scores

All verified independent and self-reported benchmark scores across standard evaluations including MMLU, GSM8K, HumanEval, and SWE-bench.

License

Open Data Commons Open Database License (ODbL) / CC-BY 4.0

Update Cadence

Real-time verification

Attribution Required

Yes (Superrank / superrank.tech)

Schema & Column Documentation (8 columns)

Column Name	Data Type	Description	Example
benchmarkName	string	Standard evaluation benchmark name	MMLU
subjectType	string	Entity category (model or agent)	model
subjectId	string	Internal identifier of evaluated entity	mod_123
score	float	Achieved evaluation score	88.7
unit	string	Measurement unit	%
selfReported	boolean	Whether the score was submitted by the vendor or third party	false
evaluator	string	Organization or harness that executed the evaluation	Epoch AI
recordedDate	ISO-8601 Date	Date the benchmark score was logged	2026-11-10