AI Compass / Quick reads
Cheaper per task is not the same as cheaper per token
A model switch only pays off if you re-measure it on your own tasks. Manufacturer figures are a starting point, not a result.
In depth: Model choice – cost per task instead of price per token
Anthropic released Claude Opus 5.5 on 22 Sept 2026. The list price is 20% below that of Opus 5: $4 per million input tokens and $20 per million output tokens. According to Anthropic, a typical task should cost 40% less. The 40% comes from the lower token price and the lower token consumption combined. Anyone comparing models should therefore measure the cost per finished task and not just read the price list.
Tool and alternative
For simple decision steps it can be even cheaper, see A model that only decides costs almost nothing.
Sources
- Anthropic, “Claude Opus 5.5”, 22 Sept 2026 (prices and the 40% are the manufacturer’s own figures)
- Jack Roberts, “Opus 5.5 vs GPT-6 Astra: 7 Real Use Cases”, 23 Sept 2026 (chapter 12:29 “division of tasks”; individual judgments, not verifiable; affiliate links in the description)
Want to try this with your team? That’s what the AI Lab is for. See the AI Lab