Gemma 2 9B

Leaderboards

Where Gemma 2 9B sits among the top 20 in each category.

By task

Scored from votes on each kind of prompt alone, on the same 1,000-point scale.

Test results

Every test Gemma 2 9B has a published result on, out of 100%, with its rank among the models that took it.

• MMLU

  #82 / 103

  Massive Multitask Language Understanding benchmark testing knowledge across 57 diverse subjects including STEM, humanities, social sciences, and professional domains

  71.3%

  Self-reported

• MATH

  #68 / 71

  MATH dataset contains 12,500 challenging competition mathematics problems from AMC 10, AMC 12, AIME, and other mathematics competitions. Each problem includes full step-by-step solutions and spans multiple difficulty levels (1-5) across seven mathematical subjects including Prealgebra, Algebra, Number Theory, Counting and Probability, Geometry, Intermediate Algebra, and Precalculus.

  36.6%

  Self-reported

• HumanEval

  #65 / 66

  A benchmark that measures functional correctness for synthesizing programs from docstrings, consisting of 164 original programming problems assessing language comprehension, algorithms, and simple mathematics

  40.2%

  Self-reported

• GSM8k

  #45 / 48

  Grade School Math 8K, a dataset of 8.5K high-quality linguistically diverse grade school math word problems requiring multi-step reasoning and elementary arithmetic operations.

  68.6%

  Self-reported

• ARC-C

  #26 / 35

  The AI2 Reasoning Challenge (ARC) Challenge Set is a multiple-choice question-answering benchmark containing grade-school level science questions that require advanced reasoning capabilities. ARC-C specifically contains questions that were answered incorrectly by both retrieval-based and word co-occurrence algorithms, making it a particularly challenging subset designed to test commonsense reasoning abilities in AI systems.

  68.4%

  Self-reported

• MBPP

  #32 / 33

  MBPP (Mostly Basic Python Problems) is a benchmark of 974 crowd-sourced Python programming problems designed to be solvable by entry-level programmers. Each problem consists of a task description, code solution, and 3 automated test cases covering programming fundamentals and standard library functionality.

  0.5%

  Self-reported

• HellaSwag

  #18 / 28

  A challenging commonsense natural language inference dataset that uses Adversarial Filtering to create questions trivial for humans (>95% accuracy) but difficult for state-of-the-art models, requiring completion of sentence endings based on physical situations and everyday activities

  81.9%

  Self-reported

• Winogrande

  #12 / 23

  WinoGrande: An Adversarial Winograd Schema Challenge at Scale. A large-scale dataset of 44,000 pronoun resolution problems designed to test machine commonsense reasoning. Uses adversarial filtering to reduce spurious biases and provides a more robust evaluation of whether AI systems truly understand commonsense or exploit statistical shortcuts. Current best AI methods achieve 59.4-79.1% accuracy, significantly below human performance of 94.0%.

  80.6%

  Self-reported

• TriviaQA

  #8 / 18

  A large-scale reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents (six per question on average) that provide high quality distant supervision for answering the questions. The dataset features relatively complex, compositional questions with considerable syntactic and lexical variability, requiring cross-sentence reasoning to find answers.

  76.6%

  Self-reported

• PIQA

  #5 / 12

  PIQA (Physical Interaction: Question Answering) is a benchmark dataset for physical commonsense reasoning in natural language. It tests AI systems' ability to answer questions requiring physical world knowledge through multiple choice questions with everyday situations, focusing on atypical solutions inspired by instructables.com. The dataset contains 21,000 multiple choice questions where models must choose the most appropriate solution for physical interactions.

  81.7%

  Self-reported

• AGIEval

  #7 / 11

  A human-centric benchmark for evaluating foundation models on standardized exams including college entrance exams (Gaokao, SAT), law school admission tests (LSAT), math competitions, lawyer qualification tests, and civil service exams. Contains 20 tasks (18 multiple-choice, 2 cloze) designed to assess understanding, knowledge, reasoning, and calculation abilities in real-world academic and professional contexts.

  52.8%

  Self-reported

• BoolQ

  #5 / 11

  BoolQ is a reading comprehension dataset for yes/no questions containing 15,942 naturally occurring examples. Each example consists of a question, passage, and boolean answer, where questions are generated in unprompted and unconstrained settings. The dataset challenges models with complex, non-factoid information requiring entailment-like inference to solve.

  84.2%

  Self-reported

• ARC-E

  #2 / 9

  ARC-E (AI2 Reasoning Challenge - Easy Set) is a subset of grade-school level, multiple-choice science questions that requires knowledge and reasoning capabilities. Part of the AI2 Reasoning Challenge dataset containing 5,197 questions that test scientific reasoning and factual knowledge. The Easy Set contains questions that are answerable by retrieval-based and word co-occurrence algorithms, making them more accessible than the Challenge Set.

  88.0%

  Self-reported

• Social IQa

  #5 / 9

  The first large-scale benchmark for commonsense reasoning about social situations. Contains 38,000 multiple choice questions probing emotional and social intelligence in everyday situations, testing commonsense understanding of social interactions and theory of mind reasoning about the implied emotions and behavior of others.

  53.4%

  Self-reported

• Natural Questions

  #3 / 7

  Natural Questions is a question answering dataset featuring real anonymized queries issued to Google search engine. It contains 307,373 training examples where annotators provide long answers (passages) and short answers (entities) from Wikipedia pages, or mark them as unanswerable.

  29.2%

  Self-reported

• BIG-Bench

  #3 / 3

  Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark consisting of 204+ tasks designed to probe large language models and extrapolate their future capabilities. It covers diverse domains including linguistics, mathematics, common-sense reasoning, biology, physics, social bias, software development, and more. The benchmark focuses on tasks believed to be beyond current language model capabilities and includes both English and non-English tasks across multiple languages.

  68.2%

  Self-reported