Kiro Gets Opus 4.8, Sonnet 5, and Admin Monitoring in GovCloud
AWS expands its Kiro IDE and CLI with two Claude models and enterprise usage tracking for government-region customers.
What AI models are now available in Kiro for AWS GovCloud (US)?
As of July 24, 2026, Kiro's IDE and CLI in the AWS GovCloud (US) regions support two Claude models: Opus 4.8 and Sonnet 5. Opus 4.8 targets complex, multi-step tasks requiring strong self-verification, while Sonnet 5 prioritizes cost-efficient agentic coding at lower credit consumption.
Model comparison
| Model | Context window | Credit multiplier | Availability status |
|---|---|---|---|
| Claude Opus 4.8 | 1M tokens | 2.2× | Generally available |
| Claude Sonnet 5 | 1M tokens | 1.3× | Experimental support |
Opus 4.8: built for long-horizon work
Opus 4.8 is positioned as the most capable model in the Opus line within Kiro. According to AWS, it plans before making edits, catches its own errors, and finds workarounds rather than stalling when it hits obstacles.
Those traits make it better suited for tasks that previously required close human oversight at each step. The 2.2× credit multiplier reflects its higher compute demands relative to Sonnet-tier models.
Sonnet 5: agentic coding at lower cost
Sonnet 5 brings reasoning and tool-use improvements that, per AWS, push its benchmark performance close to Opus 4.8 on reasoning and agentic coding tasks. Developers get a meaningful cost-performance dial: pay more for maximum accuracy with Opus 4.8, or lean on Sonnet 5 for higher throughput at the 1.3× multiplier.
Sonnet 5 carries an experimental support designation in this release, so teams with strict stability requirements should factor that into adoption decisions.
Enterprise monitoring arrives for GovCloud admins
Alongside the model additions, AWS is shipping a User Activity Monitoring feature specifically for Kiro enterprise administrators. The capability has three layers.
First, a usage dashboard shows aggregate organizational metrics at a glance. Second, per-user activity reports deliver daily CSV exports to a customer-specified S3 bucket, covering credits consumed, models used, and related telemetry useful for license optimization and audits. Third, administrators can optionally enable prompt logging, which captures user prompts and Kiro responses for compliance or debugging purposes.
All collected data lands in the customer's own AWS account. AWS states there is no additional charge beyond normal S3 storage costs.
How to access the new features
Existing Kiro users need to update their IDE or CLI to the latest version and restart it. The two models then appear in the model selector. No separate provisioning step is described in the announcement.
For organizations evaluating the monitoring feature, AWS directs administrators to the GovCloud documentation and the monitoring and tracking guide, or to their AWS account team for tailored guidance.
Why this matters for government customers
GovCloud availability means agencies and contractors operating under regulatory frameworks that require data residency in AWS GovCloud (US) regions can now use Kiro's full model lineup rather than waiting for commercial-region features to propagate. The prompt logging option in particular addresses a common compliance requirement: demonstrable records of AI-assisted work for audit purposes.