Claude Opus, Sonnet and Haiku
Anthropic's family of large language models: Opus is the most capable, Sonnet balances capability and speed, and Haiku is the fastest and cheapest.
Visit anthropic.comThe models behind my apps. Every call goes through one metered table: Opus writes, Sonnet does most of the work, Haiku sorts and triages.
Read: Meter Every AI Call From Day One- What it is
- AI models
- On my tools page
- AI and agents
- Updated
- November 2026
Where I write about Claude Opus, Sonnet and Haiku
Articles on this site that mention it.
Jul 1, 2026 · AI
Meter Every AI Call From Day One
One ledger row per model call: tokens, cost, which feature, what triggered it, and whether it changed anything. "It works" is not a metric.
Nov 12, 2026 · AI
Three Agents at Once Locked My App's AI Out for Four Hours
My app's AI shares one account limit with every agent and script I run. Three agents at once locked it out for four hours. How I run several agents now.
Nov 6, 2026 · AI
How I'm Building an AI Marketing Factory
One sentence in, a checked carousel, story set and reel out. How my AI marketing factory is built, what a week of it cost, and which calls stay mine.
Nov 5, 2026 · AI
I Built a 10-Person Content Team. None of Them Are Human.
Ten named AI teammates make content for my brands. One independent checker can fail their work, some calls stay mine, and here's what a week cost.
Nov 12, 2026 · AI
My Background AI Said OK and Did Nothing
Every background AI job in my app failed, and each one reported success. One line shape, copied into eight failure paths, hid it. How a failure says so now, and what I check instead of the tick.