Skip to content

AI benchmark index

AI Benchmark Index catalogs 264 AI benchmarks from primary sources.

Results

264 benchmarks

  • SafetyOnline

    CyberSecEval

    Evaluates cybersecurity risks from model-generated code and responses to requests that could enable cyberattacks.

  • AgentsOnline

    EmbodiedQA

    Tests agents that navigate a simulated home to gather visual evidence and answer a question.

  • RetrievalOnline

    FActScore

    Measures the factual precision of long-form generations by decomposing them into atomic claims and checking each claim against a knowledge source.

    Task count
    500

  • RetrievalOnline

    FEVER

    Tests whether systems verify Wikipedia-based claims as supported, refuted, or lacking enough evidence.

    Task count
    185,441

  • RetrievalOnline

    FinanceBench

    Evaluates open-book financial question answering against answers and evidence drawn from public company filings.

    Task count
    10,231

  • ReasoningOnline

    FinBen

    Evaluates financial language understanding and prediction across 24 tasks, including extraction, forecasting, risk, and decision making.

    Task count
    24

  • AgentsOnline

    FinGAIA

    Evaluates agents on multistep financial work involving web research, files, multimodal inputs, code, and calculations.

    Task count
    407

  • ReasoningOnline

    FinQA

    Evaluates numerical reasoning over tables and text from financial reports by requiring executable answer programs.

    Task count
    8,281

  • ReasoningOnline

    FLORES-200

    Evaluates machine translation across 200 languages with a professionally translated parallel corpus.

    Task count
    3,001

  • ReasoningOnline

    FollowBench

    Tests adherence to increasingly complex content, situation, style, format, and example constraints.

    Task count
    820

  • VoiceOnline

    Free Music Archive

    Supports music-information-retrieval evaluation with audio, metadata, features, and genre labels for Creative Commons tracks.

    Task count
    106,574

  • RetrievalOnline

    FreshQA

    Evaluates answers to questions whose facts change over time or were never present in a model’s training data.

    Task count
    600

  • MultimodalOnline

    Fréchet Video Distance

    Measures the distributional similarity of real and generated videos using learned video embeddings.

  • ScienceOnline

    GEO-Bench

    Compares pretrained models on a standardized suite of geospatial Earth-monitoring classification and segmentation datasets.

    Task count
    6

  • ReasoningOnline

    GLUE

    Combines nine sentence- and sentence-pair tasks to evaluate general-purpose English language understanding.

    Task count
    9

  • ScienceOnline

    GPQA

    Tests expert-level reasoning on graduate biology, physics, and chemistry questions designed to resist web search.

    Task count
    448

  • MathOnline

    GSM8K

    Tests multi-step arithmetic reasoning on human-written grade-school math word problems.

    Task count
    8,500

  • ReasoningOnline

    HaluEval

    Evaluates whether language models recognize hallucinated content in general queries, question answering, dialogue, and summarization.

    Task count
    35,000

  • SafetyOnline

    HarmBench

    Evaluates automated red-teaming attacks and model refusal defenses across standardized harmful behaviors.

    Task count
    510

  • ReasoningOnline

    HELM

    Evaluates foundation models across standardized scenarios with multiple accuracy, robustness, fairness, and efficiency metrics.

  • ReasoningOnline

    HotpotQA

    Tests explainable multi-hop question answering with supporting facts drawn from multiple Wikipedia articles.

  • MultimodalOnline

    HPS v2

    Evaluates whether a text-to-image model produces images people prefer for the same prompt.

    Task count
    798,000

  • CodingOnline

    HumanEval

    Tests functional correctness of generated Python functions against hidden unit tests.

    Task count
    164

  • ReasoningOnline

    Humanity’s Last Exam

    Tests frontier models on expert-written, multimodal questions spanning mathematics, science, and the humanities.

    Task count
    2,500

Cities in Global

Categories in Global

Questions

Which cities in Global have the most benchmarks?

The cities in Global with the most benchmarks are Online (264 benchmarks).

What kinds of benchmarks are in Global?

Categories in Global: Multimodal, Reasoning, Agents, Safety, and Science.

Published by . How we check each benchmark.