[BUG] "gpt-4.1" JsonReaderException for CompleteAsync

[msqishengc]

Library name and version

Azure.AI.Inference 1.0.0-beta.4

Describe the bug

I have code to guard no break when updating model version, while my code is working well with gpt-4o with Azure Inference SDK, just changing the deployment model to version gpt-4.1, it will throw error:

System.Text.Json.JsonReaderException: The input does not contain any JSON tokens. Expected the input to start with a valid JSON token, when isFinalBlock is true. LineNumber: 0 | BytePositionInLine: 0.

Stack Trace: 
ThrowHelper.ThrowJsonReaderException(Utf8JsonReader& json, ExceptionResource resource, Byte nextByte, ReadOnlySpan1 bytes) Utf8JsonReader.Read() JsonDocument.Parse(ReadOnlySpan1 utf8JsonSpan, JsonReaderOptions readerOptions, MetadataDb& database, StackRowStack& stack)
JsonDocument.Parse(ReadOnlyMemory1 utf8Json, JsonReaderOptions readerOptions, Byte[] extraRentedArrayPoolBytes, PooledByteBufferWriter extraPooledByteBufferWriter) ChatCompletions.FromResponse(Response response) <CompleteAsync>d__1.MoveNext() --- End of stack trace from previous location where exception was thrown --- ExceptionDispatchInfo.Throw() TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task) TaskAwaiter1.GetResult()
d__3.MoveNext() line 33
--- End of stack trace from previous location where exception was thrown ---
ExceptionDispatchInfo.Throw()
TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)

Here is the code:
`

[TestClass]
public class TestAzureAIInferenceWithSDK
{
Uri endpointUri = new Uri("https://.services.ai.azure.com/models");
string apiKey = "*****";
string deploymentName = "gpt-4.1"; // "gpt-4o";

 [TestMethod]
 public async Task SimpleTextPromptTest()
 {
     ChatCompletionsClient chatCompletionsClient = new ChatCompletionsClient(endpointUri, new AzureKeyCredential(apiKey));

     var requestOptions = new ChatCompletionsOptions()
     {
         Messages =
         {
             new ChatRequestSystemMessage("You are a helpful assistant"),
             new ChatRequestUserMessage("What is the result of 1+1"),
         },
         Model = deploymentName,
         MaxTokens = 800
     };

     var txtResult = await chatCompletionsClient.CompleteAsync(requestOptions);
     var rawResponse = txtResult.GetRawResponse();

     Assert.IsNotNull(rawResponse);
     Assert.AreEqual(200, rawResponse.Status);

     Assert.IsNotNull(txtResult.Value);
     Assert.IsNotNull(txtResult.Value.Content);
 }

}

`

Expected behavior

chatCompletionsClient.CompleteAsync function should work for this simple prompt for gpt-4.1 model
("What is the result of 1+1"),

Actual behavior

System.Text.Json.JsonReaderException: The input does not contain any JSON tokens. Expected the input to start with a valid JSON token, when isFinalBlock is true. LineNumber: 0 | BytePositionInLine: 0.

Reproduction Steps

1. Deploy a gpt-4.1 model (version 2025-04-14) in Azure AI foundry:

2. Create a C# console or unit test project, write sample code to ask ("What is the result of 1+1"),

`
// Replace endpointUri,apiKey and deploymentName
[TestMethod]
public async Task SimpleTextPromptTest()
{
ChatCompletionsClient chatCompletionsClient = new ChatCompletionsClient(endpointUri, new AzureKeyCredential(apiKey));

 var requestOptions = new ChatCompletionsOptions()
 {
     Messages =
     {
         new ChatRequestSystemMessage("You are a helpful assistant"),
         new ChatRequestUserMessage("What is the result of 1+1"),
     },
     Model = deploymentName,
     MaxTokens = 800
 };

 var txtResult = await chatCompletionsClient.CompleteAsync(requestOptions);
 var rawResponse = txtResult.GetRawResponse();

 Assert.IsNotNull(rawResponse);
 Assert.AreEqual(200, rawResponse.Status);

 Assert.IsNotNull(txtResult.Value);
 Assert.IsNotNull(txtResult.Value.Content);

}

`

Environment

Windows 11
Microsoft Visual Studio Enterprise 2022 (2) (64-bit) Version 17.13.6
.NETFramework 4.8 and .NET 8 also tried

[github-actions]

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @dargilco @glharper @jhakulin @trangevi.