Freemium🇺🇸
Independent benchmarking of AI models and API providers on intelligence, speed, latency and price, with image, video and coding arenas.
Best AI benchmarks and leaderboards (October 2026)
This category collects benchmarks and leaderboards that compare AI models.
In this guide:16 toolsBefore you buyHow to choose AI benchmarks and leaderboards (October 2026)Questions
16tools in this category
4countries with local strength
0can host data in the EU
How results are ranked: How we review
Results
Contact for pricing · Free plan🇺🇸
Public AI arena, formerly LMArena, that ranks models for text, code, vision, search, images and video from anonymous side-by-side votes.
Free🇭🇰🇨🇳
Chinese evaluation suite with 13,948 multiple-choice questions in 52 disciplines and four difficulty levels for foundation models.
Free🇺🇸
Collaborative benchmark of 162 English legal reasoning tasks from 40 contributors, covering contracts, statutes and court opinions.
Free🇨🇦
Evaluation set of about 12,000 expert-reviewed multiple-choice questions with ten options each, used to compare language models.
Free🇨🇳
Open-source LLM evaluation platform from Shanghai with CompassRank leaderboards, the CompassHub benchmark catalog and VLMEvalKit.
Free🇭🇰
Benchmark of 369 real computer tasks on Ubuntu, Windows and macOS that tests multimodal agents operating desktop and web apps.
Contact for pricing · Free plan🇺🇸
Scale AI's leaderboards with expert-built and partly private evaluations such as Humanity's Last Exam, SWE-Bench Pro and safety tests.
Free🇺🇸
Benchmark that tests whether AI systems resolve real GitHub issues, with Verified, Lite, Multimodal, Multilingual and Bash Only boards.
Contact for pricing🇺🇸
Independent evaluator that benchmarks AI models on professional work in law, finance, tax, healthcare, coding and other fields.
Free🇺🇸
Public leaderboard of how often language models add hallucinations when summarizing documents, scored with Vectara's open HHEM model.
FreeAvailable internationally
Open-book question answering test on public company filings from Patronus AI, with an open sample of 150 annotated cases.
FreeAvailable internationally
LLM benchmark that adds new questions regularly and scores answers against objective ground truth, without an LLM judge.
FreeAvailable internationally
Massive benchmark and toolbox for embedding and retrieval models, with more than 1,000 tasks and a public leaderboard on Hugging Face.
FreeAvailable internationally
Sierra's benchmark for AI agents that talk to users, call tools and follow policy in retail, airline, telecom and banking.
Research paperAvailable internationally
ICLR 2024 paper introducing SWE-bench: 2,294 real GitHub issues from 12 Python projects as a test bed for language models.
How to choose AI benchmarks and leaderboards (October 2026)
This category collects benchmarks and leaderboards that compare AI models. It is split into three areas: model leaderboards that rank models across many tests, arenas where people vote on anonymous answer pairs, and domain benchmarks that test specific work such as coding, customer service, legal research, search or operating a computer. Some entries are websites run by companies, others are open-source projects you can run on your own models.
Benchmarks give buyers a shared yardstick. They measure how often a model answers correctly, how fast and at what price a provider serves it, how often it invents facts in summaries or whether an agent completes a task end to end. Arenas add human preference, while private or regularly refreshed test sets try to prevent models from simply memorizing the questions. Open toolkits let your own team repeat the tests on a self-hosted or fine-tuned model.
Treat rankings as a shortlist, not a decision. Public test questions can end up in training data and inflate scores, and many benchmarks are English only or based on US law and practice. Check who runs a leaderboard and whether that company sells models or services itself. Never enter company or personal data into public arenas, since prompts may be published. Before a purchase, test the shortlisted models on your own tasks and languages.
Questions about AI benchmarks and leaderboards (October 2026)
Which leaderboard should we trust when choosing a model?
None alone. Combine a broad leaderboard, a domain benchmark close to your task and a short test with your own data. Look at how the benchmark is built, who runs it and whether the test questions are public.
Can we run benchmarks on our own models?
Yes. Open-source benchmarks such as those listed here can be run on self-hosted, fine-tuned or API models. You need compute or API budget, and some agent benchmarks need Docker or virtual machines.
Last updated: