Development tool platform Kiro announced that its IDE and command-line tools have supported access to Anthropic's latest large model, Claude Opus 4.5, to handle software engineering tasks in complex production environments. According to the official introduction, the model has achieved industry-leading results in software engineering benchmarks such as SWE-bench, and can weigh implementation options and uncertainties in a multi-system environment, making it suitable for implementing complex functions, fixing difficult defects, and supporting specification-centric development processes.
Opus 4.5 is currently available in Kiro as "experimental support" for users who log in through Google, GitHub, and AWS BuilderID, and AWS IAM Identity Center is not currently supported. The model has a billing coefficient of 2.2 times in Kiro, which is higher than 1.3 times that of Sonnet 4.5 and 0.4 times that of Haiku 4.5.
Combined with the Auto capabilities provided by Kiro, the platform recommends Opus 4.5 for difficult requirements clarification, specification and architecture design, and long-time proxy tasks, leaving daily feature development to a lower-cost model. Since it is still in the experimental stage, product details and performance may continue to be updated, and developers are encouraged to try it out in real projects and feedback on the experience through community channels.
Frequently Asked Questions
Q: What is the core content of this update?
A: Kiro has integrated Anthropic's Claude Opus 4.5 into its IDE with CLI for complex production-level software development, highlighting its leading performance on benchmarks like SWE-bench.
Q: What can Opus 4.5 do in Kiro?
A: It can be used to understand and transform large codebases, generate development specifications and architecture designs, implement complex features, fix cross-system defects, and support long-time series agents to perform multi-step development tasks.
Q: Who can currently use Opus 4.5 in Kiro?
A: This model is currently only available to users who sign in with Google, GitHub, or AWS BuilderID, and customers who sign in with AWS IAM Identity Center are not available for now.
Q: What should I pay attention to in terms of cost?
A: Opus 4.5 has a billing factor of 2.2 times in Kiro, and the cost is significantly higher than Sonnet 4.5 and Haiku 4.5, making it more suitable for high-value, complex tasks, while lower cost models can be prioritized for daily development.
Q: Is this support fully stable?
A: Officially marked Opus 4.5 as "Experimental Support", features and performance may still be adjusted, and it is recommended to continue to use it in key production scenarios with code review and existing testing and monitoring processes.
