FINAL-Bench/all-bench-leaderboard

🏆 ALL Bench Leaderboard 2026

The only AI benchmark dataset covering LLM · VLM · Agent · Image · Video · Music in a single unified file.

<p align="center"> <a href="https://huggingface.co/spaces/FINAL-Bench/all-bench-leaderboard"><img src="https://img.shields.io/badge/🏆LiveLeaderboard-ALL_Bench-6366f1?style=for-the-badge" alt="Live Leaderboard"></a> </p>

<p align="center"> <a href="https://github.com/final-bench/ALL-Bench-Leaderboard"><img src="https://img.shields.io/badge/GitHub-Repo-black?style=flat-square&logo=github" alt="GitHub"></a> <a href="https://huggingface.co/datasets/FINAL-Bench/Metacognitive"><img src="https://img.shields.io/badge/🧬FINALBench-Dataset-blueviolet?style=flat-square" alt="FINAL Bench"></a> <a href="https://huggingface.co/spaces/FINAL-Bench/Leaderboard"><img src="https://img.shields.io/badge/🧬FINALBench-Leaderboard-teal?style=flat-square" alt="FINAL Leaderboard"></a> </p>

[image]

[image]

Dataset Summary

ALL Bench Leaderboard aggregates and cross-verifies benchmark scores for 90+ AI models across 6 modalities. Every numerical score is tagged with a confidence level (cross-verified, single-source, or self-reported) and its original source. The dataset is designed for researchers, developers, and decision-makers who need a trustworthy, unified view of the AI model landscape.

[image]

[image]

[image]

What's New — v2.2.1

🏅 Union Eval ★NEW

ALL Bench's proprietary integrated benchmark. Fuses the discriminative core of 10 existing benchmarks (GPQA, AIME, HLE, MMLU-Pro, IFEval, LiveCodeBench, BFCL, ARC-AGI, SWE, FINAL Bench) into a single 1000-question pool with a season-based rotation system.

Key features:

• —100% JSON auto-graded — every question requires mandatory JSON output with verifiable fields. Zero keyword matching.
• —Fuzzy JSON matching — tolerates key name variants, fraction formats, text fallback when JSON parsing fails.
• —Season rotation — 70% new questions each season, 30% anchor questions for cross-season IRT calibration.
• —8 rounds of empirical testing — v2 (82.4%) → v3 (82.0%) → Final (79.5%) → S2 (81.8%) → S3 (75.0%) → Fuzzy (69.9/69.3%).

Key discovery: "The bottleneck in benchmarking is not question difficulty — it's grading methodology."

Empirically confirmed LLM weakness map:

• —🔴 Poetry + code cross-constraints: 18-28%
• —🔴 Complex JSON structure (10+ constraints): 0%
• —🔴 Pure series computation (Σk²/3ᵏ): 0%
• —🟢 Metacognitive reasoning (Bayes, proof errors): 95%
• —🟢 Revised science detection: 86%

Current scores (S3, 20Q sample, Fuzzy JSON):

Other v2.2 changes

• —Fair Coverage Correction: composite scoring ^0.5 → ^0.7
• —+7 FINAL Bench scores (15 total)
• —Columns sorted by fill rate
• —Model Card popup (click model name) · FINAL Bench detail popup (click Metacog score)
• —🔥 Heatmap, 💰 Price vs Performance scatter tools

Live Leaderboard

👉 [https://huggingface.co/spaces/FINAL-Bench/all-bench-leaderboard](https://huggingface.co/spaces/FINAL-Bench/all-bench-leaderboard)

Interactive features: composite ranking, dark mode, advanced search (GPQA > 90 open, price < 1), Model Finder, Head-to-Head comparison, Trust Map heatmap, Bar Race animation, Model Card popup, FINAL Bench detail popup, and downloadable Intelligence Report (PDF/DOCX).

Data Structure

data/
├── llm.jsonl           # 41 LLMs × 32 fields (incl. unionEval ★NEW)
├── vlm_flagship.jsonl  # 11 flagship VLMs × 10 benchmarks
├── agent.jsonl         # 10 agent models × 8 benchmarks
├── image.jsonl         # 10 image gen models × S/A/B/C ratings
├── video.jsonl         # 10 video gen models × S/A/B/C ratings
└── music.jsonl         # 8 music gen models × S/A/B/C ratings

LLM Field Schema

[image]

[image]

[image]

Composite Score

Score = Avg(confirmed benchmarks) × (N/10)^0.7

10 core benchmarks across the 5-Axis Intelligence Framework: Knowledge · Expert Reasoning · Abstract Reasoning · Metacognition · Execution.

v2.2 change: Exponent adjusted from 0.5 to 0.7 for fairer coverage weighting. Models with 7/10 benchmarks receive ×0.79 (was ×0.84), while 4/10 receives ×0.53 (was ×0.63).

Confidence System

Each benchmark score in the confidence object is tagged:

Example:

"Claude Opus 4.6": {
  "gpqa": { "level": "cross-verified", "source": "Anthropic + Vellum + DataCamp" },
  "arcAgi2": { "level": "cross-verified", "source": "Vellum + llm-stats + NxCode + DataCamp" },
  "metacog": { "level": "single-source", "source": "FINAL Bench dataset" },
  "unionEval": { "level": "single-source", "source": "Union Eval S3 — ALL Bench official" }
}

Usage

from datasets import load_dataset

# Load LLM data
ds = load_dataset("FINAL-Bench/ALL-Bench-Leaderboard", "llm")
df = ds["train"].to_pandas()

# Top 5 LLMs by GPQA
ranked = df.dropna(subset=["gpqa"]).sort_values("gpqa", ascending=False)
for _, m in ranked.head(5).iterrows():
    print(f"{m['name']:25s} GPQA={m['gpqa']}")

# Union Eval scores
union = df.dropna(subset=["unionEval"]).sort_values("unionEval", ascending=False)
for _, m in union.iterrows():
    print(f"{m['name']:25s} Union Eval={m['unionEval']}")

[image]

[image]

[image]

Union Eval — Integrated AI Assessment

Union Eval is ALL Bench's proprietary benchmark designed to address three fundamental problems with existing AI evaluations:

1. 1.Contamination — Public benchmarks leak into training data. Union Eval rotates 70% of questions each season.
2. 2.Single-axis measurement — AIME tests only math, IFEval only instruction-following. Union Eval integrates arithmetic, poetry constraints, metacognition, coding, calibration, and myth detection.
3. 3.Score inflation via keyword matching — Traditional rubric grading gives 100% to "well-written" answers even if content is wrong. Union Eval enforces mandatory JSON output with zero keyword matching.

Structure (S3 — 100 Questions from 1000 Pool):

Note: The 100-question dataset is not publicly released to prevent contamination. Only scores are published.

FINAL Bench — Metacognitive Benchmark

FINAL Bench measures AI self-correction ability. Error Recovery (ER) explains 94.8% of metacognitive performance variance. 15 frontier models evaluated.

• —🧬 FINAL-Bench/Metacognitive Dataset
• —🏆 FINAL-Bench/Leaderboard

Changelog

Citation

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

#AIBenchmark #LLMLeaderboard #GPT5 #Claude #Gemini #ALLBench #FINALBench #Metacognition #UnionEval #VLM #AIAgent #MultiModal #HuggingFace #ARC-AGI #AIEvaluation #VIDRAFT.net