ModelRank AI 🏆

This is an automatically updated open-source large language model leaderboard with data sourced from HuggingFace. Through this project, you can easily view and compare the performance of various large language models.

Project Features

• 🔄 Automatic Updates: Automatically fetches the latest model evaluation data from HuggingFace daily via GitHub Actions
• 📊 Complete Data: Provides comprehensive leaderboard data, including model names, parameter counts, and various evaluation scores
• 📱 Responsive Design: Supports viewing leaderboard data on various devices
• 🔍 Search and Sort: Supports searching and sorting by different metrics on the complete leaderboard page
• 📥 Data Download: Provides data in JSON and CSV formats for download

🏆 ModelRank AI Leaderboard

Last updated: 2025-10-27 01:10:03 UTC

Rank	Model	Average Score	Parameters(B)	IFEval	BBH	MATH	GPQA	MUSR	MMLU-PRO
1	MaziyarPanahi/calme-3.2-instruct-78b 📑	52.08	78.0	80.63	62.61	40.33	20.36	38.53	70.03
2	MaziyarPanahi/calme-3.1-instruct-78b 📑	51.29	78.0	81.36	62.41	39.27	19.46	36.50	68.72
3	dfurman/CalmeRys-78B-Orpo-v0.1 📑	51.23	78.0	81.63	61.92	40.63	20.02	36.37	66.80
4	MaziyarPanahi/calme-2.4-rys-78b 📑	50.77	78.0	80.11	62.16	40.71	20.36	34.57	66.69
5	huihui-ai/Qwen2.5-72B-Instruct-abliterated 📑	48.11	72.7	85.93	60.49	60.12	19.35	12.34	50.41
6	Qwen/Qwen2.5-72B-Instruct 📑	47.98	72.7	86.38	61.87	59.82	16.67	11.74	51.40
7	MaziyarPanahi/calme-2.1-qwen2.5-72b 📑	47.86	72.7	86.62	61.66	59.14	15.10	13.30	51.32
8	newsbang/Homer-v1.0-Qwen2.5-72B 📑	47.46	72.7	76.28	62.27	49.02	22.15	17.90	57.17
9	ehristoforu/qwen2.5-test-32b-it 📑	47.37	32.8	78.89	58.28	59.74	15.21	19.13	52.95
10	Saxo/Linkbricks-Horizon-AI-Avengers-V1-32B 📑	47.34	32.8	79.72	57.63	60.27	14.99	18.16	53.25
11	MaziyarPanahi/calme-2.2-qwen2.5-72b 📑	47.22	72.7	84.77	61.80	58.91	14.54	12.02	51.31
12	fluently-lm/FluentlyLM-Prinum 📑	47.22	32.8	80.90	59.48	54.00	18.23	17.26	53.42
13	JungZoona/T3Q-Qwen2.5-14B-Instruct-1M-e3 📑	47.09	0.0	73.24	65.47	28.63	22.26	38.69	54.27
14	JungZoona/T3Q-qwen2.5-14b-v1.0-e3 📑	47.09	14.8	73.24	65.47	28.63	22.26	38.69	54.27
15	zetasepic/Qwen2.5-32B-Instruct-abliterated-v2 📑	46.89	32.8	83.34	56.53	59.52	15.66	14.93	51.35
16	rubenroy/Gilgamesh-72B 📑	46.79	72.7	84.86	61.84	43.81	19.24	17.66	53.36
17	Sakalti/ultiima-72B 📑	46.77	72.7	71.40	61.10	53.55	21.92	18.12	54.51
18	CombinHorizon/zetasepic-abliteratedV2-Qwen2.5-32B-Inst-BaseMerge-TIES 📑	46.76	32.8	83.28	56.83	58.53	15.66	14.22	52.05
19	maldv/Awqward2.5-32B-Instruct 📑	46.75	32.8	82.55	57.21	62.31	12.08	13.87	52.48
20	raphgg/test-2.5-72B 📑	46.74	72.7	84.37	62.15	41.09	18.57	20.52	53.74

Complete Data

The complete leaderboard data can be viewed through the following methods:

• Online Complete Leaderboard
• JSON Format Data
• CSV Format Data

Evaluation Metrics Explanation

The leaderboard includes the following main evaluation metrics:

• Average ⬆️: Average score of all evaluations
• IFEval: Instruction following capability evaluation
• BBH: Big-Bench Hard benchmark for large language models
• MATH Lvl 5: Mathematical problem-solving capability evaluation
• GPQA: General Physics Question Answering evaluation
• MUSR: Multi-step reasoning evaluation
• MMLU-PRO: Massive Multitask Language Understanding Professional version evaluation

Local Development

Prerequisites

• Python 3.10+
• HuggingFace API token

Installation Steps

1. Clone the repository

   git clone https://github.com/chenjy16/modelrank_ai.git
   cd modelrank_ai

License

This project is open-sourced under the MIT License.

Data Source

Data is sourced from HuggingFace.