Last verified: 8 September 2026. Dollars and current names live on Claude pricing and the model tracker. This page is the pick-by-job desk, not a second rate card.
Direct Answer (8 September 2026): Default production: Sonnet 5 ($2 / $10). Hard agent / coding: Opus 5 ($5 / $25). Only if the job is worth it: Fable 5.1 ($10 / $50). Volume / smoke test: Haiku 4.5 ($1 / $5). Sonnet 4.6 at $3 / $15 and Opus 4.8 at $5 / $25 remain on the API list — do not start new work on them.
| Seat | Current name | API $ in / out | Context | Use |
|---|---|---|---|---|
| Top public | Fable 5.1 | $10 / $50 | 1M | Hardest coding and knowledge work |
| Opus | Opus 5 | $5 / $25 | 1M | Anthropic’s recommended API starting point for most workloads |
| Default | Sonnet 5 | $2 / $10 | 1M | Free / Pro default. Daily production |
| Fast | Haiku 4.5 | $1 / $5 | 200K | Volume, routing, cheap smoke tests |
Source: Anthropic API pricing and models overview.
How to pick
- New app or agent: Sonnet 5 unless you already know you need Opus 5.
- Long-horizon coding that burns Sonnet: Opus 5.
- Job that justifies 2× Opus dollars: Fable 5.1. Confirm the live ID in the console.
- Classification, extract, high QPS: Haiku 4.5.
Seats on claude.ai are a different meter. See the pricing hub. Dated ships: release history.
Track the AI tools you actually use
Live, vendor-neutral prices & limits for ChatGPT, Claude, Gemini, Perplexity and more — and we’ll email you the moment your tools change price or limits. Free, no hype.
See the live AI tracker →or set up your alerts
Leave a Reply