Anthropic Watch posted an update
AI developer MiaAIlab says a 100-HTML experiment using Fable 5.2 consumed 248.95 million tokens and cost about $512.44 at published API rates.
Why it mattersThe account says the run used 91 subagents. That gives developers a striking, if single-experiment, example of how quickly a large agent workflow can turn into a sizeable API bill. The figures are MiaAIlab’s account, not an independently audited cost breakdown.
Discuss: When testing an AI model, would you prioritise the strongest result, or the best result for the money?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.