Does livekit support gpt-realtime-1.5

Thanks for reaching out, edmon. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

Thanks for the link — I’m replying about whether LiveKit supports the OpenAI model at https://developers.openai.com/api/docs/models/gpt-realtime-1.5. Short answer: LiveKit supports the OpenAI Realtime API and uses the

gpt-realtime

model ID in its Realtime plugin, but LiveKit’s documentation does not mention a model named

gpt-realtime-1.5

. That specific model name is not confirmed as supported in the LiveKit docs. What you can do next: • If your goal is real-time audio + text interaction, use LiveKit’s OpenAI Realtime integration with the default

gpt-realtime

model ID and configure voice/temperature/modalities as needed. • If you specifically need

gpt-realtime-1.5

, could you tell me whether you have API access to that model (e.g., it appears in your OpenAI account or you see it in OpenAI docs)? With that detail I can advise whether it’s likely to work via LiveKit or suggest configuration steps to try it. I can also: • Walk through how to change the model ID in your LiveKit Realtime plugin config and test it, or • Help test behavior differences between

gpt-realtime

and any model you can access. Which would you like to do? Sources: OpenAI Realtime API plugin guide | LiveKit Documentation

hey @ancient-dog-5313, the model was released a couple of days ago, so the enum is not updated yet. But I created a PR for it However, you can still use it without that PR being merged. Look at the example voice agent: https://github.com/livekit/agents/blob/main/examples/voice_agents/weather_agent.py Update the

openai.realtime.RealtimeModel()

function with the model below:

...

class WeatherAgent(Agent):
    def __init__(self) -> None:
        super().__init__(
            instructions="You are a weather agent.",
            llm=openai.realtime.RealtimeModel(
                model="gpt-realtime-1.5",
            ),
        )

...

You can locally test it:

uv run examples/voice_agents/weather_agent.py console

Thanks — that’s great context and very helpful. I’m responding to your update about using the newly released

gpt-realtime-1.5

with LiveKit. You’re correct: even though the enum/constants in the repo haven’t been updated yet, you can already use the model by passing the model name directly to

openai.realtime.RealtimeModel(...)

, like in your example:

llm=openai.realtime.RealtimeModel(
    model="gpt-realtime-1.5",
)

And yes, you can test locally the same way shown in the voice agent example:

uv run examples/voice_agents/weather_agent.py console

If you'd like, I can help with any of the following: • Walk through testing steps and expected output when you run the example locally. • Suggest any parameters to tune (voice, temperature, modalities) for your agent. • Validate your PR or review how to update other places in your codebase where the enum might be referenced. Which of those would be most useful right now?