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

list all models available in ollama rank by low RAM/CPU usage

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Geminigemini-2.0-flash-grounding’s answer.
  1. User

    list all models available in ollama rank by low RAM/CPU usage

  2. Assistant

    Ollama offers a variety of language models tailored to different needs and hardware capabilities. For users seeking models with low RAM and CPU usage, here are some options ranked from the most resource-efficient to those requiring slightly more resources:

    1. DeepScaleR 1.5B

      • Description: A lightweight model with 1.5 billion parameters, optimized for efficient computation while maintaining impressive performance.
      • Resource Requirements: Designed for minimal resource usage, making it suitable for systems with limited RAM and CPU power.
      • Use Cases: Ideal for tasks involving mathematical reasoning and structured data processing.
    2. Orca-mini 3B

      • Description: A compact model with 3 billion parameters, offering a balance between performance and resource efficiency.
      • Resource Requirements: Can run on systems with minimal RAM, making it accessible for older or less powerful hardware.
      • Use Cases: Suitable for basic queries and simple analytical tasks.
    3. Phi-4-mini

      • Description: A lightweight open model offering significant enhancements in multilingual support, reasoning, and mathematics.
      • Resource Requirements: Excels in environments where memory and compute resources are constrained, and in scenarios where low latency is critical.
      • Use Cases: Particularly strong in reasoning tasks, especially those involving mathematics and logic.
    4. Mistral 7B

      • Description: A model with 7 billion parameters, optimized for faster inference on consumer hardware.
      • Resource Requirements: Requires at least 8 GB of RAM; performance varies based on hardware specifications.
      • Use Cases: Versatile across diverse tasks, including text generation and general-purpose AI applications.
    5. Olmo2 7B

      • Description: A 7 billion parameter model trained on up to 5 trillion tokens, competitive with other open-weight models.
      • Resource Requirements: Similar to Mistral 7B, requiring at least 8 GB of RAM.
      • Use Cases: Suitable for a wide range of tasks, with a focus on English academic benchmarks.
    6. DeepSeek-R1 7B

      • Description: A model designed for reasoning tasks, mathematics, and code operations, with 7 billion parameters.
      • Resource Requirements: Requires at least 8 GB of RAM; optimized for analytical reasoning tasks.
      • Use Cases: Excels in scenarios requiring logical analysis and structured thinking.
    7. Granite3.1-MoE 1B

      • Description: A 1 billion parameter model from IBM's Granite series, designed for low latency usage.
      • Resource Requirements: Highly efficient for on-device applications, suitable for systems with limited resources.
      • Use Cases: Effective in handling long-context tasks like summarization and question-answering.
    8. Granite-3.2 Instruct

      • Description: A sophisticated model fine-tuned for enhanced thinking capabilities, building upon the Granite-3.1 foundation.
      • Resource Requirements: Requires more resources compared to the models listed above; suitable for systems with higher RAM and CPU capabilities.
      • Use Cases: Excels in handling long-context tasks and various AI applications, including summarization and code-related tasks.

    When selecting a model, consider your specific use case and hardware capabilities to ensure optimal performance. Smaller models like DeepScaleR 1.5B and Orca-mini 3B are ideal for systems with limited resources, while models like Mistral 7B and Olmo2 7B offer a balance between performance and resource usage for more capable systems.

  3. User

    je veux un site me permettant de donné ma quantite de vram et qui me liste tout les llm qui peuvent tourner en local

  4. Assistant

    Actuellement, il n'existe pas de site spécifique où vous pouvez entrer la quantité de VRAM de votre système pour obtenir une liste des modèles de langage (LLM) pouvant fonctionner localement. Cependant, des outils comme LM Studio et Ollama facilitent l'exécution de LLM en local et peuvent vous aider à choisir des modèles adaptés à votre configuration matérielle.

    LM Studio est une application disponible sur Windows, macOS et Linux, permettant de télécharger et d'exécuter des modèles de langage open source directement sur votre machine. Elle offre une interface conviviale pour sélectionner et gérer les modèles, en tenant compte des ressources disponibles sur votre système. Par exemple, pour une configuration avec 16 Go de RAM et un GPU avec 8 Go de VRAM, des modèles comme Phi-4 ou Mistral 7B sont recommandés[1].

    Ollama est un autre outil qui permet de télécharger, gérer et exécuter des LLM directement sur votre ordinateur. Il fonctionne via la ligne de commande et prend en charge macOS, Linux et Windows. Ollama propose une bibliothèque de modèles, notamment Llama 3.2 pour les tâches textuelles et Mistral pour la génération de code. Chaque modèle fonctionne dans son propre environnement, facilitant le basculement entre différents outils d'IA pour des tâches spécifiques[2].

    Bien que ces outils ne fournissent pas directement une liste de modèles en fonction de votre VRAM, ils offrent des interfaces pour explorer et sélectionner des modèles adaptés à votre configuration. Il est recommandé de consulter la documentation de ces outils pour obtenir des recommandations spécifiques en fonction de votre matériel.

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