Amazon Web Services has expanded access to OpenAI's GPT-5.6 family on Amazon Bedrock, allowing customers in India to use the Sol, Terra and Luna models through cross-Region inference as AWS broadens availability of the models across its global cloud infrastructure.
The expansion makes GPT-5.6 accessible from AWS regions including Mumbai and Hyderabad through global cross-Region inference. Amazon Bedrock can route requests across supported commercial AWS regions based on available capacity, giving enterprises access to the models without requiring them to separately manage infrastructure across locations.
The GPT-5.6 family consists of three models positioned for different enterprise requirements. Sol is the flagship reasoning model designed for complex workloads including autonomous coding, cybersecurity research, scientific analysis and multi-step reasoning. Terra is positioned as a balanced option for general production workloads, while Luna focuses on faster and lower-cost inference for high-volume tasks such as classification, summarisation and routing.
All three models support text and image inputs and text outputs. AWS has also expanded their context windows to one million tokens, allowing businesses to process larger datasets such as complete code repositories, lengthy documents and extended histories from agentic workflows within a single request.
Developers can access the models using OpenAI's Responses API on Amazon Bedrock. AWS has additionally introduced support through its bedrock-runtime endpoint for the Responses, Converse and Chat Completions APIs, giving developers multiple options for integrating the models into existing applications.
The models also support prompt caching, which allows frequently reused portions of prompts to be stored and reused across requests. AWS says repeated context using explicit cache breakpoints can receive a 90% billing discount, a feature aimed at reducing costs for agentic workflows that repeatedly process the same information.
AWS has also lowered inference prices for two models in the family. In July, the company reduced on-demand inference prices for GPT-5.6 Luna by 80% and Terra by 20%, following similar pricing changes from OpenAI. Pricing for Sol remained unchanged.
For enterprise customers, Bedrock applies AWS identity, security and governance controls to model usage. AWS says model calls operate under Identity and Access Management policies and can be logged through AWS CloudTrail. Prompts and completions are not used to train the models or shared with the model provider, subject to AWS's stated data-handling policies.
The expansion gives Indian businesses another route to access OpenAI's latest models as cloud providers compete to support increasingly complex enterprise AI and agentic workloads.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.