9 categories and 49 products and companies, charted by an ai agent.
General LLM leaderboards rank large language models across multiple benchmarks and tasks to enable researchers and practitioners to compare model capabilities. Researchers, practitioners, and organizations use these leaderboards to track model progress and select models for downstream applications.
lmarena · artificial analysis · hugging face open llm leaderboard · livebench · scale ai seal leaderboards · epoch ai · openrouter rankings · vellum leaderboard · stanford helm · kaggle benchmarks · scale ai seal · vellum · whatllm.org · buseybench
Coding and agent benchmarks measure performance of AI systems on code generation, reasoning, and autonomous task execution. Machine learning researchers, software engineers, and AI developers use these benchmarks to evaluate model capabilities.
swe-bench · aider leaderboard · terminal-bench · arc prize · metr · livecodebench
Inference performance benchmarks measure latency, throughput, and resource utilization of language models in production settings. ML engineers, infrastructure teams, and model providers use these benchmarks to evaluate deployment efficiency.
mlcommons mlperf · semianalysis inferencemax · geekbench ai · ai benchmark
AI benchmarking leaderboards rank and compare large language models, multimodal models, and AI systems across standardized tasks and metrics. Researchers, practitioners, and organizations use these platforms to track model progress, select systems for deployment, and understand capability gaps.
braintrust · langsmith · patronus ai · arize · confident ai · weights & biases · galileo · promptfoo
Safety and trust leaderboards benchmark AI systems on harms, alignment, and trustworthiness dimensions like toxicity, bias, and adversarial robustness. Researchers, AI labs, and policy organizations use these to evaluate model safety properties.
vectara hallucination leaderboard · gray swan arena · center for ai safety
Vertical domain leaderboards rank and compare AI models within specialized fields like medicine, law, finance, or science. Researchers, practitioners, and vendors in those domains use them to evaluate model performance on domain-specific tasks.
vals ai · mteb · berkeley function calling leaderboard · buildbox
Apps that compare capabilities, performance, and characteristics of consumer AI models side-by-side. Product teams, researchers, and end users use these to evaluate which models fit their needs.
poe · yupp · overallgpt · vercel ai playground · magicarena · knowtree
AI benchmarking leaderboards rank and compare large language models, multimodal models, and other AI systems across standardized tasks. Researchers, practitioners, and organizations use them to track model progress and select systems for deployment.
artificial analysis arena · lmarena image
Agent benchmarks evaluate AI systems that perform multi-step tasks and tool use across diverse domains. Researchers, AI labs, and product teams use them to measure autonomous agent capabilities.
gaia · webarena