Note The reference leaderboard for Open LLMs! Find the best LLM for your size and precision needs, compare your models to the others! (Evaluates on ARC, HellaSwag, TruthfulQA, and MMLU)

Note Specialized leaderboard for models with coding capabilities 🖥️ (Evaluates on HumanEval and MultiPL-E)

Note Pitches chatbots against one another to compare their output quality (Evaluates on MTBench, an Elo score, and MMLU)

Note Do you want to know which model to use for which hardware? This leaderboard is for you! (Looks at the throughput of many LLMs in different hardware settings)

Note This paper introduces (among other things) the Eleuther AI Harness, a reference evaluation suite which is simple to use and quite complete!

Note The HELM paper! A super cool reference paper on the many axes to look at when creating an LLM benchmark or evaluation suite. Super exhaustive and interesting to read.

Note The BigBench paper! A bunch of tasks to evaluate edge cases and random unusual LLM capabilities. The associated benchmark has since been completed with a lot of fun crowdsourced tasks.

Note Text Embeddings benchmark across 58 tasks and 112 languages!

Note A leaderboard for tool augmented LLMs!

Note An LLM leaderboard for Chinese models on many metric axes - super complete

Note An Open LLM Leaderboard specially for Korean models by our friends at Upstage!

Note A leaderboard to evaluate the propensy of LLMs to hallucinate

Note A lot of metrics if you are interested in the propensity of LLMs to hallucinate!

Note Tests LLM API usage and calls (few models atm)

Note How likely is your LLM to help cyber attacks?

Note An aggregation of benchmarks well correlated with human preferences

Note Bias, safety, toxicity, all those things that are important to test when your chatbot actually interacts with users

Note Text to video generation leaderboard

Note Coding benchmark

Note An OCR benchmark

Note Dynamic leaderboard using complexity classes to create reasoning problems for LLMs - quite a cool one

Note Red teaming datasets success against models

Note The Open LLM Leaderboard, but for structured state models!

Note A multimodal arena!

Note An LLM leaderboard for Portuguese

Note An LLM leaderboard for Italian

Note An arena for image generation!

Note An hallucination leaderboard, focused on a different set of tasks