OpenAI’s GPT-5.6: Transforming AI with Amazon’s Bedrock Inference Engine
OpenAI has elevated the landscape of artificial intelligence by launching its most advanced family of models to date—GPT-5.6 Sol, Terra, and Luna—on Amazon’s cutting-edge Bedrock inference engine. These models are designed to meet the demanding needs of organizations leveraging AI for complex autonomous agents and product development. From cybersecurity research to genomics workflows, these models promise unrivaled intelligence and reliability.
As organizations increasingly rely on AI to handle sensitive data and unpredictable workloads, the availability of GPT-5.6 on Amazon Bedrock offers a solution that prioritizes high performance, security, and data residency. The pricing aligns with OpenAI’s proprietary pricing, contributing to existing AWS commitments.
Cutting-edge Intelligence: From Breakthrough Reasoning to Rapid Inference
With GPT-5.6, OpenAI introduces a refreshed naming system: the numeric identifier denotes the generation, while Sol, Terra, and Luna signify the models’ sustainable capability levels, each advancing individually.
- GPT-5.6 Ground: This is OpenAI’s most powerful reasoning model. Sol has set a new benchmark with an Artificial Analysis Coding Agent Index score of 80 points, surpassing the nearest model by 2.8 points while using significantly fewer tokens. It achieves a 73.5% score on the ExploitBench for cybersecurity, a substantial improvement over GPT-5.5. Sol excels in long-form professional workflows and complex reasoning tasks such as autonomous coding and drug discovery.
- GPT-5.6 Earth: Terra is ideal for daily production work, offering superior performance to GPT-5.5 at a reduced cost. It is suitable for tasks like code generation and content workflows, delivering excellent reasoning capabilities without flagship pricing.
- GPT-5.6 Moon: Luna is tailored for high-volume inference tasks, providing fast and affordable solutions for applications where latency and cost are critical.
These models are designed to complete tasks with fewer output tokens, ensuring enhanced performance per dollar spent.
An Inference Engine Designed to Scale
Amazon Bedrock’s next-generation inference engine is engineered to handle intense agent traffic, accommodating spikes in demand while maintaining isolated customer throughput. This architecture supports strict data residency requirements by keeping requests within specified AWS Regions.
GPT-5.6 on Amazon Bedrock introduces fast caching with explicit cache breakpoints, optimizing repeated context use across multi-stage workloads. Cached entries benefit from a 90% discount, remaining available for reuse for at least 30 minutes, which helps manage costs during workload scaling.
Robust Security Combined with Hardware-Enhanced Security
Recognizing the need for stronger protection, GPT-5.6 incorporates OpenAI’s most advanced security stack, combining human red teaming with automated testing. On Amazon Bedrock, this is complemented by hardware-level security, employing a zero operator access (ZOA) model. This ensures that AWS operators cannot access user prompts or completions, with all model calls recorded in AWS CloudTrail.
More Ways to Get GPT-5.6 Working on Amazon Bedrock
OpenAI has also unveiled ChatGPT Work, an agent designed for complex, multi-step tasks. The updated ChatGPT desktop app integrates Chat, Work, and Codex, offering tailored agents for diverse tasks:
- Work: Designed for extensive, multi-step tasks, capable of gathering information, utilizing the web, and creating comprehensive documents.
- Manuscript: Codex remains the go-to coding agent for developers, supporting local files, repositories, and development environments.
Users can configure applications to utilize GPT-5.6 via the Responses API on Amazon Bedrock, ensuring security and scalability backed by AWS.
Get Started Today
GPT-5.6 Sol is available in AWS Regions: US East (N. Virginia) and US East (Ohio). Terra and Luna are accessible in US East (N. Virginia), US East (Ohio), and US West (Oregon). Start using Sol, Terra, and Luna through the Amazon Bedrock console or programmatically via the Responses API. For more information, review the Amazon Bedrock documentation and visit the Amazon Bedrock product page.
About the Authors
Tanvi Girinath
Tanvi is the Product Marketing Manager for Amazon Bedrock at AWS, where she assists customers in adopting and scaling AI applications with Amazon Bedrock.
Saurabh Trikande
Saurabh is a Senior Product Manager for Amazon Bedrock at AWS. He leads initiatives to make efficient, secure, and cost-effective inferences with pioneering models available to customers at any scale.
Chris Dickens
Chris is part of the product team at OpenAI, focusing on OpenAI APIs. His work includes collaborating with AWS on Amazon Bedrock to make OpenAI’s frontier models accessible to developers.
For further details, visit the source here.
“`
