What’s Changed

AWS Bedrock now offers OpenAI’s GPT-5.6 Sol, Terra, and Luna, making it possible for developers of autonomous agents and AI products to utilize more intelligent models on high-performance, secure inference engines. GPT-5.6 completes tasks with fewer tokens than previous models, improving cost performance. (Source: OpenAI GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock)

Mechanism Details

GPT-5.6 is based on OpenAI’s new naming system, where the number represents the generation and Sol, Terra, and Luna represent a durable capability hierarchy, allowing for model selection based on workload. Additionally, the prompt cache feature has been introduced, which caches repeated contexts at cache breakpoints, resulting in only new work being charged per request, with a 90% discount applied. In terms of security, hardware-level protection, real-time misuse classifiers, and account-level enforcement measures are combined. (Source: OpenAI GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock)

Performance Characteristics

GPT-5.6 has reduced the number of tokens required to complete tasks compared to previous models, improving cost performance. It also enables capacity pooling to absorb agent traffic peaks and data localization through regional inference, allowing companies to process scalable workloads while ensuring data privacy and security. (Source: OpenAI GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock)

Summary

  • Access GPT-5.6 Sol, Terra, and Luna from the AWS Bedrock console and run AI workloads in a secure environment.
  • Utilize the prompt cache feature to reduce the cost of processing repeated contexts by 90%.
  • Enhance model safety with real-time misuse classifiers and account-level enforcement measures.