LMArena: The People's Vote

Human-preference rankings on LMArena flip between Google's Gemini, OpenAI's GPT and xAI's Grok frontier models.

The top 3

  1. LMArena's Reigning Champions: Top LLMs by User Preference: As of August 2026, Anthropic's Claude Fable 5, Claude Mythos 5, and Claude Opus 5 consistently lead the LMArena leaderboards across text, vision, and document understanding, with OpenAI's GPT-5.6 Sol and Moonshot's Kimi K3 also ranking among the top contenders.
  2. Open-Source LLMs Closing the Gap in People's Vote: The performance gap between open-source and proprietary LLMs is rapidly diminishing, with leading open models like Meta's Llama series and Chinese open-weight models achieving over 90% of closed model performance and sometimes exceeding them in specific domains by mid-2026.
  3. Task-Specific Dominance: Beyond General Chat Rankings: LMArena provides specialized sub-leaderboards for tasks like coding, vision, and long-document analysis, revealing that models such as Claude excel in coding and complex reasoning, while others like ChatGPT are preferred for general utility and multimodal interactions.

Sources

Open the full topic