The dataset viewer is taking too long to fetch the data. Try to refresh this page.
Server-side error
Error code: ClientConnectionError
Subsets and Splits
Top 100 Official Models <70
This query identifies the top 100 high-scoring, officially provided models with fewer than 70 billion parameters, offering a useful overview for comparing performance metrics.
Top 100 Official Models < 2
Identifies top-performing AI models with fewer than 20 billion parameters, offering insights into efficiency and precision in smaller models.
Top 500 Official Models by Score
Identifies top performing models based on a combined score of IFEval and MMLU-PRO metrics, filtering by official providers and parameter count, offering insights into efficient model performance.
Top 200 Official Models by Score
Discovers top high-performing models with less than 70 billion parameters, highlighting their evaluation scores and characteristics, which is valuable for model selection and optimization.
SQL Console for open-llm-leaderboard/contents
Identifies top-performing models with fewer than 70 billion parameters, combining two evaluation metrics to reveal the best balanced options.
Top 10 Official Leaderboard Models
The query identifies top 10 official providers with under 13 billion parameters, ordered by their average metric, revealing valuable insights into efficient models.
SQL Console for open-llm-leaderboard/contents
This query filters and ranks models within a specific parameter range (6-8 billion) for the LlamaForCausalLM architecture based on their average performance metric.
SQL Console for open-llm-leaderboard/contents
Retrieves entries related to chat models that are officially provided, offering a filtered view of the dataset.
SQL Console for open-llm-leaderboard/contents
The query retrieves entries marked as "Official Providers", offering basic filtering but limited analytical value.
Top 10 Official Training Data
The query retrieves a small sample of records from the 'train' dataset where the "Official Providers" flag is true, providing basic filtering with limited analytical value.