GPT-4-vision-preview first token is missing content when streaming

API Bugs
,
1

All other models send an empty string on the first token that it replies with like this:
GPT-3.5-turbo:
..."choices":[{"delta":{"role":"assistant","content":""}...

but gpt-4-vision-preview is the one model that doesn’t include this:
..."choices":[{"delta":{"role":"assistant"}...

This translates into the resulting SDKs as well; in python, for example, the proper first content token of empty string is parsed properly, however, when using gpt-4-vision-preview, it returns content=None.

Here’s some tests:
GPT-3.5-TURBO:

curl https://api.openai.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ \
  -d '{
    "model": "gpt-3.5-turbo",
    "messages": [
      {
        "role": "user",
        "content": "Hello!"
      }
    ],
    "stream": true
  }'

VISION:

curl https://api.openai.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $" \
  -d '{
    "model": "gpt-4-vision-preview",
    "messages": [
      {
        "role": "user",
        "content": "hello"
        }
    ],
    "stream" : true
  }'