GPT-5.6 Sol, Terra, and Luna now handle million-token windows with prompt caching and live web retrieval.
AWS added million-token context windows with prompt caching to the GPT-5.6 Sol, Terra, and Luna models on Amazon Bedrock. The models also gained Web Search, which pulls current information beyond their training data, folds relevant results into answers, and returns citations without a separate search provider.
Amazon Bedrock serves more than 225,000 active customers, including over 80 percent of Fortune 100 companies. AWS also offers AgentCore for building, connecting, and optimizing agents across frameworks and models, and it released the Strands Agent Harness SDK as open source.
For builders, the combination removes two common integration chores: wiring a separate search provider and splitting large documents across requests. Teams working on codebase analysis, regulatory review, or live pricing lookups can now keep more context in a single call.
AWS frames August’s releases as groundwork for agents that reason over enterprise information, run across longer time horizons, and act inside clear organizational boundaries. Watch how prompt caching economics and citation quality hold up as teams push larger working sets into production.
What matters
- AWS shipped million-token context windows with prompt caching for GPT-5.6 Sol, Terra, and Luna on Bedrock.
- Web Search gives Bedrock models current data and cites sources, removing the need for a separate provider.
- Watch how teams govern cost and freshness as agents run over longer time horizons on these systems.
Why it matters
Watch how teams govern cost and freshness as agents run over longer time horizons on these systems.
This GenAI News article was prepared in original wording using reporting and materials published by AWS Machine Learning Blog. Source reference: https://aws.amazon.com/blogs/machine-learning/icymi-what-landed-for-ai-builders-in-august-2026/.
Drafted by the GenAI News review pipeline.