Microsoft Foundry GPT-6 Sol and Luna Now GA

Introduction

Microsoft is expanding its production AI model lineup in Microsoft Foundry with the general availability of GPT-6 Sol and GPT-6 Luna, alongside GPT-6 Astra. For Azure teams building agents, this matters because model choice now becomes more practical: organizations can align reasoning depth, throughput, latency, and cost to the specific workload instead of using one model for everything.

What’s new in Microsoft Foundry

GPT-6 model lineup expands

• GPT-6 Astra remains the recommended starting point for demanding agent scenarios that require advanced reasoning, software engineering, and computer use.
• GPT-6 Sol is now GA for general-purpose production agents, complex workflows, coding, multi-step reasoning, and long-context analysis.
• GPT-6 Luna is now GA for high-volume, lower-cost tasks such as extraction, summarization, request routing, and routine customer interactions.

Deployment options for enterprise production

Microsoft says Standard deployment for Astra, Sol, and Luna is available across 28 Global regions plus US and EU Data Zones. Additional options include:

• Provisioned Throughput for Astra and Sol
• Priority Processing for Sol

This gives organizations flexibility for pay-as-you-go, reserved capacity, and lower-latency production workloads.

Pricing highlights

Microsoft is positioning the new lineup around cost per task, not just cost per token.

• GPT-6 Sol starts at $2.00 per million input tokens in Global Standard
• GPT-6 Luna starts at $0.10 per million input tokens in Global Standard
• GPT-6 Astra starts at $10.00 per million input tokens in Global Standard

Why this matters for IT and Azure administrators

For IT leaders and platform teams, the announcement is less about a single new model and more about workload optimization. Teams can now choose:

• Astra for high-stakes reasoning and decision support
• Sol for broad production agent deployments
• Luna for scaled, repetitive processing

This can improve ROI by reserving premium reasoning only for tasks that need it. Microsoft also emphasizes built-in safety controls, including guardrails, content filtering, prompt injection mitigation, identity controls, Purview policy support, and monitoring.

Recommended next steps

1. Evaluate current AI workloads and map them to Astra, Sol, or Luna based on complexity and volume.
2. Review deployment requirements for Global versus US/EU Data Zones if residency or compliance matters.
3. Test cost per task, not just token pricing, when comparing model options.
4. Assess legacy model migrations, especially if you are still using older Azure OpenAI models.
5. Use Foundry monitoring and evaluations to validate performance, safety, and operational fit before scaling.

Microsoft’s message is clear: production AI agents need more than a powerful model. With GPT-6 Astra, Sol, and Luna in Foundry, Azure customers now have a broader set of enterprise-ready options to build and scale agentic workloads.