Defined term
Foundation model
A large model pre-trained on broad data, then adapted to many downstream tasks.
A foundation model is trained on a massive corpus (text, code, images) and serves as a base for many applications via prompting, fine-tuning, or retrieval. The shift from task-specific models to foundation models is what enabled the current wave of AI-native delivery: one base model can power dozens of workflows.
When it matters
When choosing your model family for a new workflow. Foundation-model selection sets the cost, quality, and capability ceiling for everything downstream.
Real example
A bank evaluating Claude vs GPT-4 vs Gemini for a fraud-triage workflow. Eval on 1000 labelled cases: Claude scored highest on calibrated refusals (won), GPT-4 on tool-use reliability, Gemini on multilingual recall. Decision: Claude as primary, GPT-4 as fallback.
KPIs to watch
Model eval score on labelled test set (relative to top scorer), cost per call (full pricing including reasoning tokens), provider uptime SLA (>99.9%).
Related terms
LLM (Large Language Model)
A transformer-based model trained on language data to predict and generate text.
Frontier model
The leading-edge foundation models with the highest reasoning, coding, and multimodal capabilities.
Context window
The maximum number of tokens a model can process in a single request.
Transformer
The neural network architecture that powers modern LLMs, based on self-attention.
See it in action
We use this every week
Book a 30-min call and we'll walk you through how Foundation model shows up in a real engagement we're running.
Book a 30-min call