AWS: OpenAI GPT-5.6 (Sol, Terra, Luna) Now Generally Available on Amazon Bedrock
OpenAI GPT-5.6 in three variants — Sol, Terra, and Luna — is now generally available on Amazon Bedrock. Sol achieves 80 points on the Coding Agent Index, outperforms competitors by 2.8 points with significantly lower cost, and supports prompt caching with a 90% discount.
This article was generated using artificial intelligence from primary sources.
OpenAI GPT-5.6 in three variants — Sol, Terra, and Luna — has been generally available on Amazon Bedrock since July 13, 2026. This is not a model premiere (GPT-5.6 was announced on July 9th), but rather general availability within the Bedrock ecosystem with new benchmarks and prompt caching integration.
Three Models, Three Roles
GPT-5.6 comes in three variants tailored to different business needs. Sol is the flagship model for complex reasoning and agentic tasks. Terra targets production use where the balance of capability and cost is key. Luna is the fast and affordable variant for low-latency scenarios. All three variants are available to developers and businesses directly through the Amazon Bedrock API without special contracts or waiting lists.
What Does Sol Achieve on Benchmarks?
Sol achieves 80 points on the Artificial Analysis Coding Agent Index — a reference leaderboard measuring autonomous agentic capabilities in programming — which is 2.8 points above the next model on the list. It consumes less than half the output tokens and achieves roughly one-third lower costs compared to competitors at a similar performance level. On the ExploitBench cybersecurity dataset Sol reaches 73.5%, while GPT-5.5 reached 47.9% — a difference of 25.6 percentage points. On the Agent’s Last Exam test Sol achieves 53.6 points, which is 13.1 points ahead of the next model. These are differences that in practice translate into clearly visible advantages in autonomous agentic scenarios.
Prompt Caching and Pricing Policy
Prompt caching is a technique for reusing an already-processed portion of a prompt: instead of the model computationally re-processing the same context on every call, the system stores the intermediate result and charges it at a lower rate. On Bedrock this provides a 90% discount on cached input tokens with a minimum retention period of 30 minutes. Price parity with OpenAI’s first-party API is maintained for all three variants, meaning companies already using Bedrock incur no additional costs from switching platforms.
Significance for the Bedrock Ecosystem
The availability of GPT-5.6 through Amazon Bedrock means organizations that have built infrastructure on AWS can now access the most capable OpenAI model without leaving Bedrock’s management layer — including IAM policies, VPC integration, CloudWatch monitoring, and Bedrock Guardrails security controls.
Frequently Asked Questions
- What is prompt caching and what does it cost on Bedrock?
- Prompt caching is a technique for reusing an already-processed portion of a prompt without recomputing it, reducing costs. On Bedrock it provides a 90% discount on cached input tokens with a minimum retention of 30 minutes; pricing is on par with OpenAI's direct API.
- How do Sol, Terra, and Luna differ?
- Sol is the flagship model for complex reasoning, Terra is balanced for production use, and Luna is a fast and affordable variant designed for tasks requiring low latency and cost.
Sources
📬 AI news in your inbox
A daily digest built your way — pick topics, sources and cadence. One-click unsubscribe.