Claude 5.5 models reach Kiro IDE and CLI in AWS GovCloud

Anthropic’s Claude Opus 5.5 and Claude Sonnet 5.5 are now available through the Kiro IDE and command-line interface in AWS GovCloud (US) Regions. The update gives developers using Kiro in those regions access to both models, each with a 1 million-token context window.

In its availability announcement, AWS describes improvements over the preceding Opus 5 and Sonnet 5 models. The reported gains concern different measures: task-level tool and token usage for Opus, and output speed for Sonnet.

According to Kiro’s internal benchmarks, Opus 5.5 completes tasks using roughly 40% fewer tool calls and about half as many tokens as Opus 5. Those figures describe AWS’s internal testing, rather than an independently verified result. They should not be read as a guarantee of equivalent savings on every coding workload.

For Sonnet 5.5, AWS reports output that is more than 30% faster than Sonnet 5. That is an output-speed comparison, not a stated reduction in total task completion time. The distinction matters when evaluating an IDE or CLI workflow that also involves tool execution and other steps.

The announcement’s date is not firmly established by the supplied material. External corroboration describes it as October 9, 2026, while the AWS URL contains a June 2026 path. Because the supplied announcement text has no publication date, neither detail resolves the timing conclusively.

Access in Kiro carries different credit multipliers: 2.0x for Opus 5.5 and 1.3x for Sonnet 5.5. Both offer the same stated context-window capacity, so that specification alone does not distinguish the two options.

For teams assessing the update, the useful comparison is therefore workload-specific: Opus’s reported reduction in tool calls and tokens, Sonnet’s reported output-speed improvement, and each model’s credit multiplier. AWS’s token-efficiency figures do not, by themselves, establish a corresponding reduction in Kiro credit consumption or total cost.