Glossary
Foundation Model
A large-scale pre-trained model serving as the base for AI agents, providing broad capabilities before specialization or fine-tuning.
What is Foundation Model?
Foundation models like GPT-4, Claude, or Llama are trained on massive diverse datasets to develop general language understanding and reasoning. These models serve as starting points for agents, providing core capabilities that are then adapted to specific tasks through prompting, fine-tuning, or retrieval-augmented generation. Foundation models offer broad knowledge and transfer learning, reducing the data and compute needed for specialized applications.
Choosing foundation models involves evaluating capabilities, cost, latency, context window size, and licensing terms. Different foundations excel in different areas: some prioritize reasoning, others speed, others code generation. Agent performance depends heavily on foundation model quality, making model selection a critical architectural decision. Organizations may use different foundations for different agent types based on task requirements.
Example
A customer service platform builds agents on GPT-4 for complex reasoning and empathy, uses a smaller Llama model for simple FAQ responses due to lower latency and cost, and employs Claude for agents requiring strong safety and instruction-following.
How Signet addresses this
Signet's Quality and Reliability dimensions consider the foundation model as a key factor in agent capability. Agents built on proven, well-maintained foundations with strong track records score higher than those using experimental or deprecated models.
Build trust into your agents
Register your agents with Signet to receive a permanent identity and trust score.