OpenRouter launches new technology to help significantly reduce the cost of AI multi-round conversations

OpenRouter launches new technology to help significantly reduce the cost of AI multi-round conversations! In the context of the continuous development of AI dialogue technology, OpenRouter recently released two innovative technologies - Prompt Caching and Sticky Routing, aiming to significantly reduce the token cost of multiple rounds of agent dialogue. The launch of this new technology means that developers can enjoy lower costs when interacting with intelligent agents for a long time, thus promoting the popularity of intelligent agent applications. In multiple rounds of dialogue, the agent usually needs to repeatedly send the same system prompt words, tool definitions and strategy instructions. This results in developers paying multiple times for these repeated prompt words over a six-round session, even if the user only changes their input once. Through prompt word caching, OpenRouter can read repeated parts from the cache, avoiding high repeated charges. Depending on the service provider, cached reads cost only 0.1 to 0.5 times the normal input price, significantly reducing expenses. On the Claude Sonnet4.6 model, cached reads are priced at just $0.30 per million tokens, compared to the normal input price of $3.00. To ensure this mechanism continues to work across multiple conversation rounds, OpenRouter's sticky routing technology pins subsequent requests to the same provider that holds the hot cache, so developers can continue to enjoy low-cost cache reads across different conversation rounds. However, to take full advantage of this innovative technology, developers need to ensure that they maintain the same session during use. For cache effectiveness, OpenRouter recommends that users set session_id so that sticky routing is enabled immediately after the first successful request, not just after a cache hit. In addition, developers can easily confirm whether the cache is effective by viewing the usage data in the request response. With the continuous advancement of AI technology, this new technology of OpenRouter not only provides developers with a cost control tool, but also opens up new possibilities for expanding the application scenarios of intelligent agents. via AI News (author: AI Base)