Skip to content
The Chat Archive
gpt-4o-mini-search-preview· OpenAI LMArena Search Arena 24K

Could you kindly share the evaluation results for Grok-3's performance on the MMMU…

  1. User

    Could you kindly share the evaluation results for Grok-3's performance on the MMMU benchmark?

  2. Assistant

    Grok 3, xAI's latest AI model, has demonstrated strong performance on the Massive Multi-discipline Multimodal Understanding (MMMU) benchmark, achieving a score of 73.2%[1]. This benchmark evaluates models on tasks that require integrating text and images across various disciplines, including art, business, science, health, humanities, and engineering[2].

    In comparison, other leading models have achieved the following scores on the MMMU benchmark:

    • Google's Gemini 2.0: 70.7%
    • OpenAI's GPT-4o: 69.1%
    • xAI's Grok-2: 66.1%
    • Anthropic's Claude3 Opus: 59.4%[2]

    These results position Grok 3 as a competitive model in multimodal understanding tasks, showcasing its ability to process and interpret complex information involving both text and images.

  3. User

    what about grok03 thinking model?

  4. Assistant

    Grok 3's "Think" mode, introduced in February 2025, enhances its reasoning capabilities by allowing the model to engage in extended problem-solving processes, refining solutions over seconds to minutes[1].

    In evaluations, Grok 3 (Think) achieved a 93.3% score on the 2025 American Invitational Mathematics Examination (AIME), demonstrating its proficiency in mathematical reasoning[1].

    Additionally, Grok 3 (Think) attained an Elo score of 1402 in the Chatbot Arena, reflecting its strong performance in conversational contexts[1].

    These results underscore Grok 3's advanced reasoning and problem-solving capabilities, positioning it as a leading model in AI-driven reasoning tasks.

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