Skip to content
TheChatArchiveThe Chat Archive
Perplexitysonar-reasoning· PerplexityHugging Face LMArena Search Arena 24K

What is the current top-performing model on the LMArena leaderboard?

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with gpt-4o-search-preview’s answer.
  1. User

    What is the current top-performing model on the LMArena leaderboard?

  2. Perplexity Assistant

    The current top-performing model on the LMArena leaderboard is DeepSeek-R1, which holds an Arena Elo rating of 1361 as of March 17, 2025[1]. This model leads the "Best Open LM" category, followed closely by Gemma-3-27B-it (1339 Elo) and DeepSeek-V3 (1319 Elo)[1].

    Key details about the leaderboard:

    • Benchmarking methodology: Uses a Bradley-Terry model for centralized rating calculations based on pairwise comparisons[1][4].
    • Recent updates: The leaderboard incorporates over 2.7 million user votes and evaluates models across multiple benchmarks like MMLU and Arena-Hard-Auto[1].
    • Competitors: Other notable models include Qwen2.5-72B-Instruct (1257 Elo) and Llama-3.3-70B-Instruct (1256 Elo)[1].

    For specialized categories like web development, Claude 3.7 Sonnet currently leads the WebDev Arena[12].

  3. User

    wrong this is not the last update

  4. Perplexity Assistant

    The current top-performing model on the LMArena leaderboard is DeepSeek-R1, holding an Arena Elo rating of 1361 as of March 17, 2025[1]. It leads the "Best Open LM" category, followed by Gemma-3-27B-it (1339 Elo) and DeepSeek-V3 (1319 Elo)[1].

    Key details:

    • Benchmarking: The leaderboard uses a Bradley-Terry model for centralized rating calculations based on pairwise comparisons, incorporating over 2.7 million user votes[1][4].
    • Specialized categories: In the WebDev Arena (focused on web app development), Claude 3.7 Sonnet currently leads with an Arena Score of 1169.19[3].

    Other notable models include Qwen2.5-72B-Instruct (1257 Elo) and Llama-3.3-70B-Instruct (1256 Elo)[1].

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