Live Webinar On: Building AI-First Financial InstitutionsRegister Now
    AI Glossary · Foundations

    AI Model

    A trained mathematical function that maps inputs to outputs. The core building block of AI.

    Category · Foundations3 min readUpdated August 2026

    What is AI Model?

    n AI model is a trained mathematical function that maps inputs to outputs, whether text to text, images to labels, or speech to transcript. Modern AI models are typically neural networks trained on large datasets. Large language models are AI models trained to process and generate natural language. Enterprise deployments increasingly favour private, on-premise AI models to preserve data sovereignty.

    Choosing an AI model for an enterprise deployment involves trade-offs across four dimensions: capability, cost, latency, and data governance. A frontier model like GPT-4o or Claude 3.5 Sonnet offers the highest raw capability but requires sending data to an external API, which is a non-starter for most regulated industries. Open-weight models like Llama 3.1 or Mistral offer deployable-on-premise capability, increasingly competitive with frontier models for specific tasks, with full data sovereignty. Distilled and quantized variants of these models further reduce compute requirements for deployment on standard enterprise hardware.

    The model is not the application. This is the most common misconception in enterprise AI projects. An LLM out of the box is a general-purpose text predictor. The application — the prompt design, the retrieval layer, the tool integrations, the guardrails, the audit logging — is what makes it useful for a specific enterprise workflow. Model selection matters, but organisations that focus exclusively on which model to use and under-invest in the surrounding application stack consistently achieve poor production outcomes.

    Also known as: Machine Learning Model, Foundation Model

    Key Points

    Key Points

    • Core idea

      Frontier API models offer peak capability but require data to leave your infrastructure. Open-weight on-premise models offer sovereignty. Fine-tuned on-premise models can close most of the capability gap for specific tasks.

    • Why it matters

      The model is one component. The prompt layer, retrieval system, tool connectors, guardrails, and audit infrastructure determine whether an AI deployment succeeds in production.

    • Enterprise use

      A 7B parameter model runs on a single A100 GPU. A 70B parameter model requires multiple GPUs. Quantization and distillation let you run larger-capability models on constrained enterprise hardware.

    How It Works

    How AI Model works

    1. Define the purpose, inputs, and success criteria that AI Model must support.

    2. Apply AI Model in the relevant workflow while recording its inputs, configuration, and outputs.

    3. Evaluate the result against representative data, operational constraints, and human review before expanding production use.

    How Fluid AI Uses This

    Private AI models on customer infrastructure.

    Fluid AI supports both open-source and proprietary AI models deployed on customer infrastructure. Models can run air-gapped, on-premise, or in private cloud. Model weights and training data stay inside your perimeter.

    Explore Deployment Options

    Topics Covered

    • enterprise AI model selection
    • on-premise LLM deployment
    • private AI model
    • AI model vs AI application
    • open-weight models enterprise
    • foundation model deployment
    • AI model governance
    • AI model cost latency trade-off
    Continue Exploring

    Related terms in Foundations.

    Want to see how Fluid AI uses this in production?

    Book a 30-minute session with our enterprise AI team.

    Book a Demo