Claude Opus 4.8 available in Microsoft Foundry
- Maxime Hiez
- Foundry
- 02 Jun, 2026
Introduction
Microsoft made Claude Opus 4.8 available in Microsoft Foundry in preview, a few days after its general release by Anthropic on May 28, 2026. The model is accessible via Global Standard deployment in two Azure regions and integrates into the existing Foundry ecosystem via the Claude Messages API. Access is restricted to Enterprise and MCA-E subscriptions.
Check the Claude Opus 4.8 article HERE.
Deployment and regions
The model is identified by the identifier claude-opus-4-8 in Foundry. It is deployed via the Global Standard type and is available in the following regions :
- East US 2
- Sweden Central
The Foundry project or hub must be located in one of these two regions for the deployment to be possible. The context window is 1M tokens, with a maximum output of 128K tokens. The deployment procedure is identical to other Claude models available in Foundry : Models from partners in the ai.azure.com portal, then model selection and deployment configuration.
Access conditions and quotas
Access to Claude models in Foundry is restricted to Enterprise and MCA-E subscriptions. Other subscription types are assigned a default quota of 0 RPM and 0 TPM, making the model unusable without an explicit increase request.
For eligible subscriptions, the default limits are :
- 2000 requests per minute (RPM)
- 2 000 000 tokens per minute (TPM)
A quota increase request form is available via the Azure portal for organizations that exceed these limits.
info
Adaptive thinking and effort control
Opus 4.8 introduces adaptive thinking in Foundry, an evolution of the extended thinking mode (extended thinking). Unlike previous versions that supported “enabled”, “disabled”, and “adaptive” types for the “thinking” parameter, Opus 4.8 only supports “adaptive” and “disabled”. The model decides on its own the intensity of reasoning required based on the complexity of the request.
The “effort” parameter controls the quality/cost tradeoff. Opus 4.8 supports the following levels :
- “low”, “medium”, “high”, “max”
- “xhigh” (equivalent to “max”, kept for backward compatibility)
Both parameters can be used together or independently. For teams migrating from Opus 4.7, a migration guide is available in the Anthropic documentation.
Conclusion
The availability of Claude Opus 4.8 in Microsoft Foundry strengthens the offering of top-tier third-party models accessible via the Azure ecosystem. The restriction to Enterprise and MCA-E subscriptions remains a friction point for teams operating on standard subscriptions. For organizations already deployed on Opus 4.7 in Foundry, migrating to Opus 4.8 requires no infrastructure change, but validating behaviors related to adaptive thinking is recommended before moving to production.
Sources
Microsoft Learn - Deploy Claude models in Foundry
Microsoft Learn - Region availability for models
Did you enjoy this post ? If you have any questions, comments or suggestions, please feel free to send me a message from the contact form.
Don’t forget to follow us and share this post.
