Home / Blogs / Agentic AI
Agentic AI · 15 Sep 2026 · 2 min read
Claude
Claude is a family of large language models made by Anthropic, available as a web and mobile assistant, a developer API, and through cloud marketplaces. Tiers run Opus, Sonnet and Haiku, from most capable to fastest.
Made by Anthropic
at a glance
- Maker
- Anthropic
- Category
- Model family
- Tiers
- Opus, Sonnet, Haiku
- Interfaces
- Web, mobile, desktop
- Developer API
- Yes
- Related tools
- Claude Code, Cowork
- Status
- Generally available
Check current pricing with the vendor. Rates change often and we do not republish them.
what it does
One family, three sizes
The tiers trade capability against cost and speed. Opus is the most capable and the slowest, Sonnet is the general-purpose default, Haiku is the fastest and cheapest. The same models are reachable from the consumer apps and from the API, so a team can prototype in the interface and then ship against the endpoint without changing model.
choosing a tier
Start at Sonnet, move outward
The common mistake is standardising on the largest tier before measuring. Start at the middle tier, then move up only for the specific steps that fail and down for the high-volume steps that do not need it. In agentic workloads the planner often needs the top tier while the tool-formatting steps do not.
what to check
Before you standardise on it
-
What are the data retention and training terms on your specific contract tier, in writing?
-
Direct contract or cloud marketplace? It changes who you escalate to when something breaks.
-
What does a tier change cost in engineering time, not just in rate per token?
-
What is your migration path to a different provider, and has anyone tested it?
firms that build on it
Who we have seen deliver on Claude
Aulric Labs
Tredence
Marlowe AI
Questions people ask
Who makes Claude?
Anthropic, a public benefit corporation founded in 2021 and based in San Francisco.
What are the Claude tiers?
Opus, Sonnet and Haiku, ordered from most capable to fastest and cheapest. All three are available via the API and the apps.
Which tier should we use for agents?
Usually the top tier for the planner and a cheaper tier for the mechanical steps. Standardising on one tier for the whole pipeline is the expensive mistake.
Does Claude have a developer API?
Yes, with the same model family available through it as through the consumer and team interfaces.
Read this next
Agent evaluation becomes a separate line item in RFPs
Keep reading
Agentic AI
Agent evaluation becomes a separate line item in RFPs
Updated 16 Sep 2026: added Aulric Labs opening its harness to non-platform customers
Agentic AI
How to tell a real agent from a wrapped prompt
Almost every firm in our Agentic AI area now uses the word agent. Some are shipping…
Data Science
How leading enterprises are using supply chain analytics to stay ahead of disruption
Discover how leading enterprises use supply chain analytics, predictive forecasting, and real-time visibility to stay ahead…
AI Consulting
How healthcare leaders are choosing AI consulting partners that actually deliver
Learn the five signals that separate a real healthcare AI consulting partner from an expensive pilot,…