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

which LLM is the best LLM under 36B NOW?

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

    which LLM is the best LLM under 36B NOW?

  2. Assistant

    As of April 2025, several large language models (LLMs) with fewer than 36 billion active parameters have demonstrated strong performance across various benchmarks. Here are some notable models:

    DBRX
    Developed by Mosaic ML and Databricks, DBRX is a mixture-of-experts transformer model with a total of 132 billion parameters, of which 36 billion are active per token. Released in March 2024, it outperformed other open-source models like Meta's LLaMA 2 and Mistral AI's Mixtral in language understanding, programming ability, and mathematics[1].

    Mixtral 8x22B
    Introduced by Mistral AI in April 2024, Mixtral 8x22B is a sparse Mixture-of-Experts model with 141 billion total parameters, utilizing 39 billion active parameters per token. It has been recognized for its efficiency and performance, offering a favorable performance-to-cost ratio[2].

    Command R
    Released by Cohere in March 2024, Command R is a scalable LLM with 35 billion parameters. It supports ten languages and offers a 128,000-token context length. The model excels in retrieval-augmented generation, code-related tasks, and reasoning[2].

    Llama 3 (8B and 70B variants)
    Meta's Llama 3 series, released in April 2024, includes models with 8 billion and 70 billion parameters. The 8B variant offers competitive performance in a compact size, while the 70B model provides enhanced capabilities[3].

    Falcon 7B
    Part of the Falcon series introduced in November 2023, Falcon 7B is a 7-billion-parameter model trained on diverse, high-quality corpora. It offers efficient performance and is suitable for various applications[4].

    Each of these models has unique strengths, and the choice of the "best" LLM under 36 billion active parameters depends on specific use cases, performance requirements, and resource constraints.

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