Just shipped a big refresh of our open leaderboard Space: BenchHub Leaderboards
BenchHub Leaderboards - a Hugging Face Space by runbenchhub
The idea: stop comparing models apples-to-oranges. Every model is scored on the same data and the same metric, reproducibly — so the ranking actually means something.
What’s live:
Vision — classification, segmentation (mIoU), detection (mAP), mono + stereo depth, optical flow
Audio — ESC-50 classification, ASR (WER)
NLP — text classification, SQuAD QA, NER
95 boards · 700+ submissions · free to submit (sign in with 
, GitHub, or Google and run a one-line client on your predictions).
Browse the standings here; the full explorer + sample-by-sample comparison lives at runbenchhub.com. Feedback + benchmark requests very welcome! 
1 Like