What a task costs on Opus 5.5 (claude.dev)

🤖 AI Summary
The recent announcement of Opus 5.5 reveals that it offers a significant reduction in token costs compared to its predecessor, Opus 5, making AI model usage more efficient for developers. Specifically, input tokens are now priced at $4 per million (20% lower), output tokens at $20 (also 20% lower), and cache reads are priced dramatically lower at $0.20 per million (60% reduction). This update allows for more efficient workflows as it emphasizes the importance of both the number of tokens used and the model's operational efficiency, with guidance on how to optimize tasks to minimize token costs. For the AI/ML community, the implications of Opus 5.5 are profound, particularly in terms of task efficiency. The model is designed to reduce the number of turns and the associated costs by encouraging practices like checking work through tests, which avert costly retries. Its strategic pricing structure incentivizes the use of cache reads and careful modeling of input efforts, potentially resulting in savings of up to 60% on input costs for longer sessions driven by cached data. Users are encouraged to explore customized settings to further enhance cost-effectiveness, framing Opus 5.5 as a versatile tool for developers focused on building features, performing migrations, or executing complex tasks with greater cost-efficiency.
Loading comments...
loading comments...