GPT-5.6 Sol enters routed coding workflows in Kiro and Replit

OpenAI says the GPT-5.6 family, including Sol, Terra, and Luna, is now available in Kiro, while a separate OpenAI account of Replit Free Mode says tasks requiring advanced reasoning can be routed from Luna to Sol without losing project context. These are product-integration claims, not a new release claim for Sol, and they describe where a pre-existing model family is being surfaced to developers.

The Kiro announcement frames the integration around spec-driven development rather than a bare chat interface. OpenAI says Kiro turns high-level intent into requirements, technical designs, and executable tasks, then gives the model codebase and team-standard context; it also describes review checkpoints and property-based testing. For teams already working from written specifications, that is a more concrete operating model than asking a coding assistant to infer intent from an isolated prompt.

The evidence for performance needs careful separation by model tier. OpenAI reports a Terminal-Bench 2.1 result for GPT-5.6 Terra in Kiro, not Sol, and attributes the reported cost reduction to testing in that environment. Replit, meanwhile, presents Luna as the Free Mode model and Sol as the route for tasks that need more advanced reasoning. Neither announcement supplies an independent Sol-versus-Luna quality comparison or a public account of the routing threshold, so the product architecture should not be read as a benchmark result.

For technical users, the practical change is that model choice is increasingly embedded in the coding product rather than exposed only as a single fixed selection. Kiro presents the full GPT-5.6 family inside its workflow, while Replit describes a cost-oriented default that can escalate to Sol. Availability and capability still vary by host: Amazon's India Bedrock announcement covers Terra and Luna, not Sol, illustrating that a family-level integration does not establish identical regional or platform access for every tier. Teams evaluating Sol should therefore test their own repository, review gates, latency, and spending constraints rather than infer them from partner announcements.