Azure OpenAI Model: gpt-4.1 context window exceeded with way less than 1M tokens

Question

Hello!

I'm having trouble using a large context window with gpt-4.1.

gpt-4.1 is known for having a 1M token context window.

It is described as such in azure docs:
https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models?tabs=global-standard%2Cstandard-chat-completions#capabilities

When I try to send something like 300k tokens, I get the following error:

openai.BadRequestError: Error code: 400 - {'error': {'message': 'Your input exceeds the context window of this model. Please adjust your input and try again.', 'type': 'invalid_request_error', 'param': 'input', 'code': 'context_length_exceeded'}}

Below is a sample code I'm using to send the message (python sdk):

response: Response = client.responses.create(
            model=AZURE_AI_OPENAI_DEPLOYMENT_NAME,
            max_output_tokens=32768,
            instructions="system prompt with around 2k tokens",
            input="large message with around 300k tokens",
            store=False,
        )

api_version of my client is 2025-04-01-preview.

I'd like to know what is the actual maximum context window of gpt-4.1 within Azure.

My model is deployed at West US 3 with Global Standard type and a rate limit around 2M tokens per minute. When this rate is exceeded, I get another message, so I don't think that's what's preventing me from using the full gpt-4.1 context window.

Thanks a lot for any clarification on this matter.

Answer 1

Hi Thiago Almeida,

Even though GPT-4.1 supports up to 1 million tokens, in Azure OpenAI, the full 1M-token context is only available on deployments of the gpt-4-1106-preview or newer models where explicitly enabled, and only for certain SKUs and regions.

Your West US 3 deployment and model configuration likely do not support the 1M context, even if you're using gpt-4.1. In most regions and configurations, the max context is still 128k tokens (some older ones are 32k).

Understanding GPT-4.1's Context Window in Azure OpenAI

While GPT-4.1 is advertised to support a context window of up to 1 million tokens, this capability is not universally available across all Azure OpenAI deployments. The actual context window limit can vary based on several factors:

1.Model Variant: Different variants of GPT-4.1 (e.g., gpt-4-1106-preview, gpt-4-32k) have different context window capacities. For instance, gpt-4-32k supports up to 32,768 tokens.

2.Deployment Region: Certain regions may not yet support the full 1 million token context window. It's essential to verify the capabilities available in your specific deployment region.

3.API Version: The API version used can influence the features and limitations accessible in your deployment.

In your case, deploying GPT-4.1 in the West US 3 region with the 2025-04-01-preview API version may not currently support the full 1 million token context window, which could explain the errors encountered when sending inputs around 300k tokens.

You can refer to the following official Azure documentation:

·   Azure OpenAI Service Models: Provides an overview of available models and their capabilities.

o   Azure OpenAI Models

·   Quotas and Limits: Details the quotas and limitations associated with Azure OpenAI services.

o   Azure OpenAI Quotas and Limits

·   Transparency Note: Offers insights into the limitations and considerations when using GPT-4.1.

o   Azure OpenAI Transparency Note

Hope this helps. If you have any follow-up questions, please let me know. I would be happy to help.

**

Please do not forget to "Accept the answer” and “up-vote” wherever the information provided helps you, this can be beneficial to other community members.

Thank you!