Choose a model
Models supply inference capabilities to agents and applications. Treat model selection as a workload decision rather than a permanent architecture choice.
Compare models using:
- Task quality against a fixed evaluation set
- Context and modality support
- Latency and throughput
- Input, output, and hosting cost
- Security and data-retention terms
- Region and data-residency options
- Hosted, private, or open-weight deployment requirements
Keep tools, policies, context assembly, and evaluations in the AI Agent Harness so you can change models without rebuilding the workflow.
Browse AI models →Last modified on September 15, 2026
