Exact context window for gpt-4.1 (2025-04-14) on Global Standard deployment — factors not covered by documentation

Question

I have a gpt-4.1 deployment (model version 2025-04-14) with the following exact configuration:

• Deployment type: Global Standard
• Resource pricing tier: S0
• Region: italynorth
• API version: 2024-02-15-preview

The official documentation (learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models) lists three context window figures for this model: 1,047,576, 300,000 (standard deployments), and 128,000 (provisioned managed and batch deployments) — without stating which applies to Global Standard specifically, nor how the resource pricing tier (S0 vs higher) factors in.

I've also seen it confirmed in another support thread that the actual usable context window can vary based on model variant, deployment region, and API version — factors not reflected in the documentation table above.

Given my exact configuration listed above, could you confirm:

1. Which exact context window (max total tokens: prompt + completion) currently applies?
2. Whether the S0 pricing tier itself imposes any additional limit on context window, separate from the Global Standard deployment type?
3. Whether upgrading to a newer API version would change this limit?
4. Whether this figure is documented anywhere explicitly per region/tier/API version, or only obtainable by contacting support / testing empirically?

Answer 1

For gpt-4.1 (2025-04-14), the documented context window depends on deployment category, but the provided documentation does not explicitly map that limit to the exact combination of Global Standard + italynorth + S0 + API version 2024-02-15-preview.

Supported facts from the documentation:

1. gpt-4.1 supports these documented context-window figures:
   • 1,047,576
   • 300,000 for standard deployments
   • 128,000 for provisioned managed and batch deployments
2. A separate model page lists:
   • 1,047,576
   • 128,000 for provisioned managed deployments
   • 300,000 for batch deployments
3. The documentation also states a known issue for all GPT-4.1 series models: large tool or function definitions exceeding 300,000 tokens can fail even if the 1 million token context limit is not reached.
4. No provided documentation states that the S0 pricing tier changes the context window for gpt-4.1.
5. No provided documentation states that changing the API version changes the context window for gpt-4.1.
6. No provided documentation gives an explicit per-region / pricing tier / API version context-window matrix for gpt-4.1.
7. The model summary table in the provided context does not list gpt-4.1 availability for italynorth.

Because of that, the exact current limit for the specific configuration cannot be confirmed from the provided documentation alone.

---

References:

• Foundry Models sold by Azure (azure-openai)
• Azure OpenAI in Azure AI Foundry Models