runbenchhub/leaderboards

🏆 BenchHub Leaderboards

A read-only mirror of public leaderboard standings from [runbenchhub.com](https://runbenchhub.com) — an open, multi-modal model benchmarking platform. Browse the standings here; run your own model and submit on BenchHub (free).

What's benchmarked

Live boards across a growing set of domains, each with real models scored on the same eval set:

• —LLM / Reasoning — MMLU (14k MCQ, 57 subjects), HellaSwag, ARC-Challenge, GSM8K — one pinned prompt, scored by exact match, zero-shot, so every model is evaluated identically
• —Vision — Image Classification, Semantic Segmentation (mIoU), Object Detection (mAP), Optical Flow, Monocular & Stereo Depth Estimation, Point Tracking, Image Captioning
• —Audio — Audio Classification (ESC-50), Automatic Speech Recognition (WER)
• —NLP — Text Classification, Extractive Question Answering (SQuAD), Named-Entity Recognition (CoNLL-2003), Visual Question Answering, Translation

Submit your model

Every "Submit" button here deep-links to that leaderboard's submission page on BenchHub. Submitting is free — sign in with GitHub, Google, or 🤗 Hugging Face, run the one-line client on your predictions, and see where you rank.

This Space holds no ground-truth samples, predictions, or submission UI — it reads a derived standings dataset (HF_RESULTS_REPO) that BenchHub publishes. The full interactive experience (per-sample explorer, GT visualizations, model comparison) lives on [runbenchhub.com](https://runbenchhub.com).