About AI Benchmark

Purpose

AI Benchmark provides a centralized resource for comparing large language model (LLM) performance across standardized benchmarks. Our goal is to help researchers, developers, and organizations make informed decisions about model selection.

Benchmarks

We track performance on widely-used evaluation benchmarks including:

• MMLU - Massive Multitask Language Understanding
• HumanEval - Code generation evaluation
• GSM8K - Grade school math word problems
• HellaSwag - Commonsense natural language inference
• MATH - Competition mathematics problems
• BBH - BIG-Bench Hard reasoning tasks

Data Sources

Benchmark scores are compiled from official model documentation, academic papers, and verified third-party evaluations. We prioritize accuracy and transparency in our data collection.

API Access

Public API endpoints are available for programmatic access:

• GET /api/benchmarks - List all benchmarks
• GET /api/models - List all models
• GET /api/leaderboard/:benchmarkId - Get leaderboard for a benchmark
• GET /api/scores/:modelId - Get all scores for a model