# Global MMLU Lite (Kaggle) > Kaggle reports an average across 16 language dimensions of Global MMLU Lite, with separate culturally sensitive and culturally agnostic results. Canonical page: https://benchlm.ai/benchmarks/global-mmlu-lite-kaggle - Category: [Multilingual](/multilingual) - Last updated: Source retrieved 2026-10-20T15:19:04.746Z Evaluation results published by [Cohere Labs](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite). undefined The source dataset has 400 questions per language, split evenly between culturally sensitive and culturally agnostic samples, using human translations or post-edits. Its current dataset card covers 23 languages; this Kaggle v1 board instead labels its published Average as an average of 16 language dimensions. We preserve that published average and the individual source metrics without recomputing a broader score. This Kaggle protocol is separate from Artificial Analysis Global-MMLU-Lite and the full Global-MMLU dataset. The source response includes task version 1 for Average and task version 2 for language and cultural-sensitivity metrics. Unspecified model settings remain unknown. These results do not enter weighted rankings. These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference. - [Kaggle leaderboard and Apache 2.0 declaration](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) - [Cohere Labs Global-MMLU-Lite dataset card](https://huggingface.co/datasets/CohereLabs/Global-MMLU-Lite) - [Global-MMLU paper](https://arxiv.org/abs/2412.03304) ## About Global MMLU Lite (Kaggle) - Year: 2026 - Tasks: 400 test questions per language; source average of 16 language dimensions - Format: Source-published percentage score - Difficulty: Multilingual multiple-choice knowledge The source dataset has 400 questions per language, split evenly between culturally sensitive and culturally agnostic samples, using human translations or post-edits. Its current dataset card covers 23 languages; this Kaggle v1 board instead labels its published Average as an average of 16 language dimensions. We preserve that published average and the individual source metrics without recomputing a broader score. Global MMLU Lite (Kaggle) is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula. ## Leaderboard (45 models) | Rank | Model | Creator | Score | |------|-------|---------|-------| | 1 | [Claude Opus 5](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 95.7% | | 2 | [Gemini 3.5 Flash](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 95.4% | | 3 | [Gemini 4 Argon](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 95.4% | | 4 | [Claude Opus 5.5](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 95.4% | | 5 | [Gemini 3.8 Flash](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 95.3% | | 6 | [Gemini 3.7 Flash](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 94.9% | | 7 | [Claude Opus 4.8](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 94.9% | | 8 | [Gemini 3 Pro Preview](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 94.5% | | 9 | [Claude Opus 4.1](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 94.3% | | 10 | [Gemini 3.6 Flash](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 94.3% | | 11 | [GPT-5.6 Sol](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 94.1% | | 12 | [GPT-6 Sol](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 94.1% | | 13 | [Grok 4.5](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 93.3% | | 14 | [Gemini 2.5 Pro](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 93.2% | | 15 | [GPT-5.6 Terra](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 92.7% | | 16 | [Claude Sonnet 5.5](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 92.6% | | 17 | [GPT-6 Astra](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 92.6% | | 18 | [GPT-6 Luna](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 92.0% | | 19 | [Gemini 2.5 Flash](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 91.5% | | 20 | [Gemini 2.5 Flash Preview](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 90.9% | | 21 | [Gemma 4 31B](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 90.7% | | 22 | [Claude Sonnet 4](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 90.6% | | 23 | [Gemini 3.5 Flash-Lite](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 89.5% | | 24 | [GPT-5](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 89.0% | | 25 | [Grok 4](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 88.8% | | 26 | [Qwen 3 235B A22B Instruct 2506](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 88.0% | | 27 | [GPT-4.1](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 87.5% | | 28 | [Claude Haiku 5.5](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 87.3% | | 29 | [o4 mini](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 87.0% | | 30 | [Gemma 4 26B A4B](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 85.7% | | 31 | [Command A ](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 83.9% | | 32 | [GPT-5.6 Luna](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 83.1% | | 33 | [Claude 3.7 Sonnet](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 80.8% | | 34 | [deepseek-v3.1](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 80.4% | | 35 | [GPT-6.1 Sol](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 78.9% | | 36 | [Mistral Small 3.1](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 78.5% | | 37 | [o3 mini](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 78.0% | | 38 | [Gemma 3 27B](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 76.3% | | 39 | [Granite 4.0 Small](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 75.0% | | 40 | [Aya Expanse 32B](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 73.5% | | 41 | [DeepSeek-R1](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 67.4% | | 42 | [Grok 3 Mini](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 67.3% | | 43 | [Gemma 3 4B](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 65.1% | | 44 | [Claude 3.5 Haiku](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 61.1% | | 45 | [Mistral Medium 3](https://www.kaggle.com/benchmarks/cohere-labs/global-mmlu-lite) | — | 55.1% | ## FAQ ### What does Global MMLU Lite (Kaggle) measure? Kaggle reports an average across 16 language dimensions of Global MMLU Lite, with separate culturally sensitive and culturally agnostic results. ### Which source row has the highest reported Global MMLU Lite (Kaggle) value? Claude Opus 5 has the highest reported Global MMLU Lite (Kaggle) value at 95.7%. These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference. ### How many models are evaluated on Global MMLU Lite (Kaggle)? The Source retrieved 2026-10-20T15:19:04.746Z contains 45 AI models. ### Does Global MMLU Lite (Kaggle) affect BenchLM's overall score? Not directly. Global MMLU Lite (Kaggle) is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.