README.md

metadata

title: All Bench Leaderboard
emoji: 🏆🏆🏆
colorFrom: indigo
colorTo: pink
sdk: static
pinned: false
license: apache-2.0
short_description: Benchmarking metrics to 90+ leading Generation AI models
models:
  - Qwen/Qwen3.5-122B-A10B
  - Qwen/Qwen3.5-27B
  - Qwen/Qwen3.5-35B-A3B
  - Qwen/Qwen3.5-9B
  - Qwen/Qwen3.5-4B
  - Qwen/Qwen3-Next-80B-A3B-Thinking
  - deepseek-ai/DeepSeek-V3
  - deepseek-ai/DeepSeek-R1
  - zai-org/GLM-5
  - meta-llama/Llama-4-Scout-17B-16E-Instruct
  - meta-llama/Llama-4-Maverick-17B-128E-Instruct
  - microsoft/phi-4
  - upstage/Solar-Open-100B
  - K-intelligence/Midm-2.0-Base-Instruct
  - Nanbeige/Nanbeige4.1-3B
  - MiniMaxAI/MiniMax-M2.5
  - stepfun-ai/Step-3.5-Flash
  - OpenGVLab/InternVL3-78B
  - Qwen/Qwen2.5-VL-72B-Instruct
  - Qwen/Qwen3-VL-30B-A3B
  - black-forest-labs/FLUX.1-dev
  - stabilityai/stable-diffusion-3.5-large
  - Lightricks/LTX-Video
  - facebook/musicgen-large
  - facebook/jasco-chords-drums-melody-1B
tags:
  - leaderboard
  - benchmark
  - evaluation
  - llm
  - gpt
  - claude
  - gemini
  - deepseek
  - qwen
  - korean-ai
  - sovereign-ai
  - arc-agi-2
  - gpqa
  - mmlu-pro
  - swe-bench
  - hle
  - metacognitive
  - final-bench
  - aime
  - reasoning
  - multimodal
  - open-source
  - closed-source
  - ai-comparison
  - model-ranking
  - k-ai
  - exaone
  - solar
  - llm-benchmark
  - agi

🏆 ALL Bench Leaderboard 2026 — Unified Multi-Modal AI Benchmark

The only leaderboard comparing LLM · VLM · Agent · Image · Video · Music across 6 modalities in a single view.

Why ALL Bench?

Most leaderboards only compare LLM text scores. ALL Bench cross-verifies 42 LLMs + 11 VLMs + 10 Agents + 28 generative models and presents them in a single unified view. Every score carries a ✓✓ confidence badge — hover to see the exact source.

v2.1 Features

Feature	Description
📊 LLM Leaderboard	42 models × 27 metrics with composite 5-axis scoring
👁 VLM Benchmark	11 flagship models (Gemini · GPT · Claude · InternVL3 · Kimi) + 5 edge models
🤖 Agent Bench	OSWorld · BrowseComp · Terminal-Bench 2.0 · SWE-Pro
🖼🎬🎵 Generative AI	Image 10 · Video 10 · Music 8 model comparison
📄 Intelligence Report	Auto-generated Executive Summary with PDF/DOCX download
✓✓ Confidence System	Cross-verified · Single source · Self-reported — 3-tier badges
🔍 Analysis Tools	Model Finder · Head-to-Head · Trust Map · Bar Race
📡 Free API	8 Gradio endpoints, no auth required

Composite Score Formula

Score = Avg(verified benchmarks) × √(N/10)

10 Core Benchmarks: MMLU-Pro · GPQA · AIME · HLE · ARC-AGI-2 · FINAL Bench (Metacognition) · SWE-Pro · BFCL · IFEval · LiveCodeBench

5-Axis Framework: Knowledge · Expert Reasoning · Abstract Reasoning · Metacognition · Execution

FINAL Bench — Metacognitive Benchmark

FINAL Bench measures an AI model's self-correction ability. Error Recovery (ER) explains 94.8% of metacognitive performance variance.

• 🧬 Dataset: FINAL-Bench/Metacognitive
• 🏆 Leaderboard: FINAL-Bench/Leaderboard
• 📰 Featured in: Seoul Shinmun · Asia Economy · IT Chosun (2026.02.27)

API Usage

from gradio_client import Client
client = Client("VIDRAFT/ALL-Bench")

# Get all LLM data
llm = client.predict(api_name="/get_llm_data")

# Search models
results = client.predict("Claude", "llm", api_name="/search_models")

# Get everything at once
all_data = client.predict(api_name="/get_all_data")

Data Files

File	Contents
llm.json	42 LLM models (27 metrics each)
vlm.json	VLM benchmarks (34 benchmarks)
agent.json	Agent benchmarks (8 benchmarks)
image.json	10 image generation models
video.json	10 video generation models
music.json	8 music generation models

Citation

@misc{allbench2026,
    title={ALL Bench Leaderboard 2026: Unified Multi-Modal AI Evaluation},
    author={ALL Bench Team},
    year={2026},
    url={https://huggingface.co/spaces/VIDRAFT/ALL-Bench}
}

---

ALL Bench Leaderboard v2.1 · Updated 2026.03.08

#AIBenchmark #LLMLeaderboard #GPT5 #Claude #Gemini #ALLBench #FINALBench #Metacognition #VLM #AIAgent #MultiModal #HuggingFace #OpenSource #AIEvaluation #DeepLearning #ARC-AGI