I have a Visual Studio subscription through my employer that includes $150 per month in Azure credits. I'm trying to deploy several AI models in Azure AI Foundry for testing, but most of the newer models are unavailable. Global Standard lets me deploy some older or lower-tier options, such as GPT-4o, embedding models, MaaS, Qwen3, and GPT-OSS, but I can't deploy the GPT-5.4, GPT-5.5, or GPT-5.6-Luna variants.
My quota increase request for East US 2 was rejected with a message saying that the region couldn't expand capacity. I also tried West US and Canada, with the same result. Does a Visual Studio subscription count as a trial or otherwise have lower model-access limits? Are there specific regions where quota increases for these models are actually available, or should I contact support instead of continuing to guess?
4 Answers
The newest models may start with zero quota for these subscriptions, so an increase request can be rejected even when the region itself is normally supported. Existing deployments may continue working, which is why older models such as GPT-5.4 can remain usable while newer versions cannot be deployed.
Some people see Visual Studio subscriptions as having trial-like restrictions, but the subscription is generally treated as pay-as-you-go with monthly credits. Model availability and quota can still depend on subscription history, capacity, and the specific model. For newer models, opening a support request may work better than repeatedly submitting self-service quota increases.
This seems to be a real limitation for some Visual Studio credit subscriptions. Moving the deployment to a separate, standard Azure subscription gave one user access to many more models, although the service didn’t clearly explain that model selection could be restricted under Visual Studio credits.
There’s also some confusion about what Azure credits cover. Certain third-party marketplace services may not qualify, but some Azure AI model deployments do charge against the monthly credit balance. So credit eligibility alone probably doesn’t explain every rejected deployment; subscription quota and model capacity appear to matter too.
I’ve been able to deploy at least one newer model and have the usage deducted from the monthly credits, so it doesn’t seem like all AI Foundry models are excluded.
That makes sense. I’ll try opening a support request rather than continuing to switch regions.