{"benchmark_id":"multi-if","benchmark_name":"Multi-IF","benchmark_description":"Multi-IF benchmarks LLMs on multi-turn and multilingual instruction following. It expands upon IFEval by incorporating multi-turn sequences and translating English prompts into 7 other languages, resulting in 4,501 multilingual conversations with three turns each. The benchmark reveals that current leading LLMs struggle with maintaining accuracy in multi-turn instructions and shows higher error rates for non-Latin script languages.","max_score":1.0,"categories":["chat","language","reasoning","structured_output","instruction_following","communication"],"modality":"text","total_models":23,"entries":[{"rank":1,"model_id":"qwen3-235b-a22b-thinking-2507","model_name":"Qwen3-235B-A22B-Thinking-2507","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.806,"normalized_score":0.806,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-07-25","announcement_date":"2025-07-25","multimodal":false,"param_count":235000000000,"is_new":false},{"rank":2,"model_id":"lfm-2.5-2.6b","model_name":"LFM2.5-2.6B","organization_name":"Liquid AI","organization_id":"liquid","benchmark_score":0.8007,"normalized_score":0.8007,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-08-04","announcement_date":"2026-08-04","multimodal":false,"param_count":2690000000,"is_new":false},{"rank":3,"model_id":"o3-mini","model_name":"o3-mini","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.795,"normalized_score":0.795,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-01-30","announcement_date":"2025-01-30","multimodal":false,"param_count":null,"is_new":false},{"rank":4,"model_id":"qwen3-vl-235b-a22b-thinking","model_name":"Qwen3 VL 235B A22B Thinking","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.791,"normalized_score":0.791,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":236000000000,"is_new":false},{"rank":5,"model_id":"qwen3-vl-32b-thinking","model_name":"Qwen3 VL 32B Thinking","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.78,"normalized_score":0.78,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":33000000000,"is_new":false},{"rank":6,"model_id":"qwen3-next-80b-a3b-thinking","model_name":"Qwen3-Next-80B-A3B-Thinking","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.778,"normalized_score":0.778,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-10","announcement_date":"2025-09-10","multimodal":false,"param_count":80000000000,"is_new":false},{"rank":7,"model_id":"qwen3-235b-a22b-instruct-2507","model_name":"Qwen3-235B-A22B-Instruct-2507","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.775,"normalized_score":0.775,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-07-22","announcement_date":"2025-07-22","multimodal":false,"param_count":235000000000,"is_new":false},{"rank":8,"model_id":"qwen3-vl-235b-a22b-instruct","model_name":"Qwen3 VL 235B A22B Instruct","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.763,"normalized_score":0.763,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":236000000000,"is_new":false},{"rank":9,"model_id":"qwen3-next-80b-a3b-instruct","model_name":"Qwen3-Next-80B-A3B-Instruct","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.758,"normalized_score":0.758,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-10","announcement_date":"2025-09-10","multimodal":false,"param_count":80000000000,"is_new":false},{"rank":10,"model_id":"qwen3-vl-8b-thinking","model_name":"Qwen3 VL 8B Thinking","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.751,"normalized_score":0.751,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":9000000000,"is_new":false},{"rank":11,"model_id":"qwen3-vl-8b-instruct","model_name":"Qwen3 VL 8B Instruct","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.751,"normalized_score":0.751,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":9000000000,"is_new":false},{"rank":12,"model_id":"qwen3-vl-4b-thinking","model_name":"Qwen3 VL 4B Thinking","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.736,"normalized_score":0.736,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":0.1,"output_cost_per_million":1.0,"speed_rps":null,"context_window":262144,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":4000000000,"is_new":false},{"rank":13,"model_id":"qwen3-vl-30b-a3b-thinking","model_name":"Qwen3 VL 30B A3B Thinking","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.73,"normalized_score":0.73,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":31000000000,"is_new":false},{"rank":14,"model_id":"qwen3-30b-a3b","model_name":"Qwen3 30B A3B","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.722,"normalized_score":0.722,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":0.1,"output_cost_per_million":0.3,"speed_rps":82.57,"context_window":128000,"release_date":"2025-04-29","announcement_date":"2025-04-29","multimodal":false,"param_count":30500000000,"is_new":false},{"rank":15,"model_id":"qwen3-vl-32b-instruct","model_name":"Qwen3 VL 32B Instruct","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.72,"normalized_score":0.72,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":33000000000,"is_new":false},{"rank":16,"model_id":"gpt-4.1-2025-04-14","model_name":"GPT-4.1","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.708,"normalized_score":0.708,"verified":false,"self_reported":true,"provider_id":"openai","input_cost_per_million":2.0,"output_cost_per_million":8.0,"speed_rps":100.0,"context_window":1047576,"release_date":"2025-04-14","announcement_date":"2025-04-14","multimodal":true,"param_count":null,"is_new":false},{"rank":17,"model_id":"gpt-4.5","model_name":"GPT-4.5","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.708,"normalized_score":0.708,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-02-27","announcement_date":"2025-02-27","multimodal":true,"param_count":null,"is_new":false},{"rank":18,"model_id":"gpt-4.1-mini-2025-04-14","model_name":"GPT-4.1 mini","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.67,"normalized_score":0.67,"verified":false,"self_reported":true,"provider_id":"openai","input_cost_per_million":0.4,"output_cost_per_million":1.6,"speed_rps":150.0,"context_window":1047576,"release_date":"2025-04-14","announcement_date":"2025-04-14","multimodal":true,"param_count":null,"is_new":false},{"rank":19,"model_id":"qwen3-vl-30b-a3b-instruct","model_name":"Qwen3 VL 30B A3B Instruct","organization_name":"Alibaba Cloud / Qwen Team","organization_id":"qwen","benchmark_score":0.661,"normalized_score":0.661,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2025-09-22","announcement_date":"2025-09-22","multimodal":true,"param_count":31000000000,"is_new":false},{"rank":20,"model_id":"gpt-4o-2024-08-06","model_name":"GPT-4o","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.609,"normalized_score":0.609,"verified":false,"self_reported":true,"provider_id":"openai","input_cost_per_million":2.5,"output_cost_per_million":10.0,"speed_rps":132.0,"context_window":128000,"release_date":"2024-08-06","announcement_date":"2024-08-06","multimodal":true,"param_count":null,"is_new":false},{"rank":21,"model_id":"lfm-2.5-vl-3b","model_name":"LFM2.5-VL-3B","organization_name":"Liquid AI","organization_id":"liquid","benchmark_score":0.594,"normalized_score":0.594,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-08-12","announcement_date":"2026-08-12","multimodal":true,"param_count":3123483888,"is_new":false},{"rank":22,"model_id":"gpt-4.1-nano-2025-04-14","model_name":"GPT-4.1 nano","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.572,"normalized_score":0.572,"verified":false,"self_reported":true,"provider_id":"openai","input_cost_per_million":0.1,"output_cost_per_million":0.4,"speed_rps":200.0,"context_window":1047576,"release_date":"2025-04-14","announcement_date":"2025-04-14","multimodal":true,"param_count":null,"is_new":false},{"rank":23,"model_id":"north-micro-vision-instruct","model_name":"North Micro Vision Instruct","organization_name":"Cohere","organization_id":"cohere","benchmark_score":0.373,"normalized_score":0.373,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-08-12","announcement_date":"2026-08-12","multimodal":true,"param_count":2400000000,"is_new":false}]}