{"benchmark_id":"drop","name":"DROP","parent_benchmark":null,"categories":["math","reasoning"],"modality":"text","multilingual":false,"max_score":1.0,"language":"en","description":"DROP (Discrete Reasoning Over Paragraphs) is a reading comprehension benchmark requiring discrete reasoning over paragraph content. It contains crowdsourced, adversarially-created questions that require resolving references and performing discrete operations like addition, counting, or sorting, demanding comprehensive paragraph understanding beyond paraphrase-and-entity-typing shortcuts.","paper_link":"https://arxiv.org/abs/1903.00161","implementation_link":null,"verified":false,"created_at":"2026-05-07T16:53:22.863541+00:00","updated_at":"2026-07-05T18:27:47.110803+00:00","statistics":{"total_models":30,"average_score":0.7245866666666666,"min_score":0.286,"max_score":0.916,"score_stddev":0.16071501385421624,"verified_count":0,"self_reported_count":29},"child_benchmarks":[],"linked_dataset":null,"models":[{"rank":1,"model_id":"deepseek-v3","model_name":"DeepSeek-V3","organization_id":"deepseek","organization_name":"DeepSeek","organization_country":"CN","score":0.916,"normalized_score":0.916,"verified":false,"self_reported":true,"self_reported_source":"https://github.com/deepseek-ai/DeepSeek-V3","analysis_method":"3-shot F1","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-12-25","param_count":671000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":2,"model_id":"claude-3-5-sonnet-20241022","model_name":"Claude 3.5 Sonnet","organization_id":"anthropic","organization_name":"Anthropic","organization_country":"US","score":0.871,"normalized_score":0.871,"verified":false,"self_reported":true,"self_reported_source":"https://www.anthropic.com/news/3-5-models-and-computer-use","analysis_method":"3-shot F1 Score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-10-22","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":2,"model_id":"claude-3-5-sonnet-20240620","model_name":"Claude 3.5 Sonnet","organization_id":"anthropic","organization_name":"Anthropic","organization_country":"US","score":0.871,"normalized_score":0.871,"verified":false,"self_reported":true,"self_reported_source":"https://www.anthropic.com/news/claude-3-5-sonnet","analysis_method":"3-shot F1 Score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-06-21","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":4,"model_id":"mimo-v2.5-pro","model_name":"MiMo-V2.5-Pro","organization_id":"xiaomi","organization_name":"Xiaomi","organization_country":"CN","score":0.863,"normalized_score":0.863,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/XiaomiMiMo/MiMo-V2.5-Pro","analysis_method":"Base model, 3-shot","verification_date":null,"provider_id":"xiaomi","input_cost_per_million":0.435,"output_cost_per_million":0.87,"context_window":1048576,"announcement_date":"2026-04-27","param_count":1023244718976,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":"Xiaomi","best_throughput":null,"throughput_provider":"Xiaomi","context_provider":"DeepInfra"},{"rank":5,"model_id":"gpt-4-turbo-2024-04-09","model_name":"GPT-4 Turbo","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.86,"normalized_score":0.86,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/hello-gpt-4o/","analysis_method":"Reading comprehension & arithmetic (f1 score)","verification_date":null,"provider_id":"openai","input_cost_per_million":10.0,"output_cost_per_million":30.0,"context_window":128000,"announcement_date":"2024-04-09","param_count":null,"is_open_source":false,"is_new":false,"best_latency":0.5,"latency_provider":"OpenAI","best_throughput":100.0,"throughput_provider":"OpenAI","context_provider":"OpenAI"},{"rank":6,"model_id":"nova-pro","model_name":"Nova Pro","organization_id":"amazon","organization_name":"Amazon","organization_country":"US","score":0.854,"normalized_score":0.854,"verified":false,"self_reported":true,"self_reported_source":"https://www.amazon.science/publications/the-amazon-nova-family-of-models-technical-report-and-model-card","analysis_method":"0-shot","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-11-20","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":7,"model_id":"llama-3.1-405b-instruct","model_name":"Llama 3.1 405B Instruct","organization_id":"meta","organization_name":"Meta","organization_country":"US","score":0.848,"normalized_score":0.848,"verified":false,"self_reported":true,"self_reported_source":"https://arxiv.org/pdf/2407.21783","analysis_method":"0-shot","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-07-23","param_count":405000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":8,"model_id":"gpt-4o-2024-05-13","model_name":"GPT-4o","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.834,"normalized_score":0.834,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/blog/gpt-4o","analysis_method":"F1 Score","verification_date":null,"provider_id":"azure","input_cost_per_million":2.5,"output_cost_per_million":10.0,"context_window":128000,"announcement_date":"2024-05-13","param_count":null,"is_open_source":false,"is_new":false,"best_latency":0.5,"latency_provider":"OpenAI","best_throughput":100.0,"throughput_provider":"OpenAI","context_provider":"Azure"},{"rank":9,"model_id":"claude-3-opus-20240229","model_name":"Claude 3 Opus","organization_id":"anthropic","organization_name":"Anthropic","organization_country":"US","score":0.831,"normalized_score":0.831,"verified":false,"self_reported":true,"self_reported_source":"https://www.anthropic.com/news/claude-3-family","analysis_method":"3-shot, F1 Score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-02-29","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":9,"model_id":"claude-3-5-haiku-20241022","model_name":"Claude 3.5 Haiku","organization_id":"anthropic","organization_name":"Anthropic","organization_country":"US","score":0.831,"normalized_score":0.831,"verified":false,"self_reported":true,"self_reported_source":"https://www.anthropic.com/news/claude-3-5-haiku","analysis_method":"3-shot F1 Score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-10-22","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":11,"model_id":"gpt-4-0613","model_name":"GPT-4","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.809,"normalized_score":0.809,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/research/gpt-4","analysis_method":"3-shot, Reading comprehension & arithmetic (f1 score)","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2023-06-13","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":12,"model_id":"nova-lite","model_name":"Nova Lite","organization_id":"amazon","organization_name":"Amazon","organization_country":"US","score":0.802,"normalized_score":0.802,"verified":false,"self_reported":true,"self_reported_source":"https://www.amazon.science/publications/the-amazon-nova-family-of-models-technical-report-and-model-card","analysis_method":"0-shot","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-11-20","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":13,"model_id":"gpt-4o-mini-2024-07-18","model_name":"GPT-4o mini","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.797,"normalized_score":0.797,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/blog/gpt-4o-mini-announcement","analysis_method":"F1 Score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-07-18","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":14,"model_id":"llama-3.1-70b-instruct","model_name":"Llama 3.1 70B Instruct","organization_id":"meta","organization_name":"Meta","organization_country":"US","score":0.796,"normalized_score":0.796,"verified":false,"self_reported":true,"self_reported_source":"https://arxiv.org/pdf/2407.21783","analysis_method":"0-shot","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-07-23","param_count":70000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":15,"model_id":"nova-micro","model_name":"Nova Micro","organization_id":"amazon","organization_name":"Amazon","organization_country":"US","score":0.793,"normalized_score":0.793,"verified":false,"self_reported":true,"self_reported_source":"https://www.amazon.science/publications/the-amazon-nova-family-of-models-technical-report-and-model-card","analysis_method":"6-shot Chain-of-Thought","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-11-20","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":16,"model_id":"longcat-flash-chat","model_name":"LongCat-Flash-Chat","organization_id":"meituan","organization_name":"Meituan","organization_country":"CN","score":0.7906,"normalized_score":0.7906,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/meituan-longcat/LongCat-Flash-Chat","analysis_method":"F1 score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-08-29","param_count":560000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":17,"model_id":"claude-3-sonnet-20240229","model_name":"Claude 3 Sonnet","organization_id":"anthropic","organization_name":"Anthropic","organization_country":"US","score":0.789,"normalized_score":0.789,"verified":false,"self_reported":true,"self_reported_source":"https://www.anthropic.com/news/claude-3-family","analysis_method":"3-shot, F1 score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-02-29","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":18,"model_id":"claude-3-haiku-20240307","model_name":"Claude 3 Haiku","organization_id":"anthropic","organization_name":"Anthropic","organization_country":"US","score":0.784,"normalized_score":0.784,"verified":false,"self_reported":true,"self_reported_source":"https://www.anthropic.com/news/claude-3-haiku","analysis_method":"3-shot, F1 score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-03-13","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":19,"model_id":"phi-4","model_name":"Phi 4","organization_id":"microsoft","organization_name":"Microsoft","organization_country":"US","score":0.755,"normalized_score":0.755,"verified":false,"self_reported":true,"self_reported_source":"https://arxiv.org/pdf/2412.08905","analysis_method":"simple-evals","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-12-12","param_count":14700000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":20,"model_id":"gemini-1.5-pro","model_name":"Gemini 1.5 Pro","organization_id":"google","organization_name":"Google","organization_country":"US","score":0.749,"normalized_score":0.749,"verified":false,"self_reported":true,"self_reported_source":"https://arxiv.org/pdf/2403.05530","analysis_method":"Variable shots","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-05-01","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":21,"model_id":"gpt-3.5-turbo-0125","model_name":"GPT-3.5 Turbo","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.702,"normalized_score":0.702,"verified":false,"self_reported":false,"self_reported_source":"https://example.com/benchmark-image","analysis_method":"Accuracy","verification_date":null,"provider_id":"openai","input_cost_per_million":0.5,"output_cost_per_million":1.5,"context_window":16385,"announcement_date":"2023-03-21","param_count":null,"is_open_source":false,"is_new":false,"best_latency":0.5,"latency_provider":"OpenAI","best_throughput":100.0,"throughput_provider":"OpenAI","context_provider":"OpenAI"},{"rank":22,"model_id":"gemma-3n-e4b-it-litert-preview","model_name":"Gemma 3n E4B Instructed LiteRT Preview","organization_id":"google","organization_name":"Google","organization_country":"US","score":0.608,"normalized_score":0.608,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/google/gemma-3n-E4B-it-litert-preview","analysis_method":"1-shot Token F1 score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-05-20","param_count":1910000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":22,"model_id":"gemma-3n-e4b","model_name":"Gemma 3n E4B","organization_id":"google","organization_name":"Google","organization_country":"US","score":0.608,"normalized_score":0.608,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/google/gemma-3n-E4B","analysis_method":"Token F1 score. 1-shot.","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-06-26","param_count":8000000000,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":24,"model_id":"llama-3.1-8b-instruct","model_name":"Llama 3.1 8B Instruct","organization_id":"meta","organization_name":"Meta","organization_country":"US","score":0.595,"normalized_score":0.595,"verified":false,"self_reported":true,"self_reported_source":"https://arxiv.org/pdf/2407.21783","analysis_method":"0-shot","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2024-07-23","param_count":8000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":25,"model_id":"granite-3.3-8b-instruct","model_name":"Granite 3.3 8B Instruct","organization_id":"ibm","organization_name":"IBM","organization_country":"US","score":0.5936,"normalized_score":0.5936,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/ibm-granite/granite-3.3-8b-instruct","analysis_method":"OLMES (Modified implementation)","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-04-16","param_count":8000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":26,"model_id":"gemma-3n-e2b","model_name":"Gemma 3n E2B","organization_id":"google","organization_name":"Google","organization_country":"US","score":0.539,"normalized_score":0.539,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/google/gemma-3n-E2B","analysis_method":"Token F1 score. 1-shot.","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-06-26","param_count":8000000000,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":26,"model_id":"gemma-3n-e2b-it-litert-preview","model_name":"Gemma 3n E2B Instructed LiteRT (Preview)","organization_id":"google","organization_name":"Google","organization_country":"US","score":0.539,"normalized_score":0.539,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/google/gemma-3n-E2B-it-litert-preview","analysis_method":"1-shot Token F1 score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-05-20","param_count":1910000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":28,"model_id":"granite-4.0-tiny-preview","model_name":"IBM Granite 4.0 Tiny Preview","organization_id":"ibm","organization_name":"IBM","organization_country":"US","score":0.462,"normalized_score":0.462,"verified":false,"self_reported":true,"self_reported_source":"https://www.ibm.com/new/announcements/ibm-granite-4-0-tiny-preview-sneak-peek","analysis_method":"Score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-05-02","param_count":7000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":29,"model_id":"granite-3.3-8b-base","model_name":"Granite 3.3 8B Base","organization_id":"ibm","organization_name":"IBM","organization_country":"US","score":0.3614,"normalized_score":0.3614,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/ibm-granite/granite-3.3-8b-base","analysis_method":"Score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-04-16","param_count":8170000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":30,"model_id":"ernie-4.5","model_name":"ERNIE 4.5","organization_id":"baidu","organization_name":"Baidu","organization_country":"CN","score":0.286,"normalized_score":0.286,"verified":false,"self_reported":true,"self_reported_source":"https://cloud.baidu.com/doc/WENXINWORKSHOP/s/clntwmv7t","analysis_method":null,"verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-06-25","param_count":21000000000,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null}]}