Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-search-preview· OpenAIHugging Face LMArena Search Arena 24K

what is the top ten model in livebench now?

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Perplexitysonar-pro-high’s answer.
  1. User

    what is the top ten model in livebench now?

  2. Assistant

    As of the latest available data from November 2024, the top ten models on LiveBench are:

    1. o1-preview-2024-09-12: Leading with a global average score of 64.74, this model excels in language tasks and data analysis, making it a top choice for general-purpose applications[1].

    2. claude-3-5-sonnet-20241022: With a global average of 56.68, it stands out in coding tasks, achieving a coding average of 67.13, indicating strong capabilities in programming-related applications[1].

    3. o1-mini-2024-09-12: Scoring 56.01 globally, this model leads in reasoning tasks with a score of 72.33, suggesting a focus on logical inference capabilities[1].

    4. gemini-exp-1121: With a global average of 56.01, it performs well in instruction following (IF score of 80.15) and mathematics (62.75), indicating strengths in directive-driven tasks and numerical reasoning[1].

    5. step-2-16k-202411: Holding a global average of 55.09, this model excels in instruction following with an IF score of 79.88, making it suitable for rule-based tasks[1].

    6. gpt-4o-2024-11-20: Achieving a global average of 54.88, it offers balanced performance across various tasks without excelling in a specific area, making it suitable for general-purpose use[1].

    7. mistral-large-2411: With a global average of 54.74, this model shows consistent performance across multiple categories, indicating reliability in diverse applications[1].

    8. nova-pro-v1: Scoring 54.50 globally, it demonstrates strong capabilities in data analysis with a score of 63.97, making it a solid choice for analytical tasks[1].

    9. llama-3.1-nemotron-70b-instruct: With a global average of 54.21, this model performs well in language tasks, achieving a language score of 68.72, indicating advanced language generation and comprehension[1].

    10. qwen-2.5-coder-32b-instruct: Holding a global average of 54.19, it shows balanced performance across various tasks, making it suitable for general-purpose applications[1].

    Please note that these rankings are based on data from November 2024, and newer evaluations may have been conducted since then. For the most current information, I recommend visiting the official LiveBench website or consulting their latest publications.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.