Pre-deployed models
LiveHub provides several pre-deployed Large Language Models (LLMs) that can be used in AI Agents.
The following textual models are currently provided:
-
OpenAI models:
-
gpt-4o
-
gpt-4o-mini
-
gpt-4.1
-
gpt-4.1-mini
-
gpt-4.1-nano
-
gpt-5
-
gpt-5-mini
-
gpt-5-nano
-
gpt-5.2-chat
-
gpt-5.4
-
gpt-5.4-mini
-
gpt-5.4-nano
-
-
Google models:
-
gemini-2.0-flash
-
gemini-2.0-flash-lite
-
gemini-2.5-flash
-
gemini-2.5-flash-lite
-
gemini-3-flash
-
gemini-3.1-flash-lite
-
Each LLM model has its own strengths and weaknesses, excelling in some tasks and underperforming in others, and they also differ in speed and operational cost depending on their architecture and optimization.
Choosing between the models
For typical voice agent use cases, we recommend starting with one of the following models:
-
gpt-4o-mini
-
gpt-4.1-mini
-
gpt-5.4-mini
-
gemini-2.5-flash
-
gemini-3.1-flash-lite
These models offer good balance of capabilities, latency and cost efficiency.
Experiment with different models, as a model’s performance is highly dependent on your prompts.
If you see that your agent struggles with following complex instructions, consider switching to larger models, for example, gpt-4o, gpt-4.1, or gpt-5.2-chat.
Speech-to-speech models
Speech-to-speech models communicate with user using audio modality and completely bypass speech-to-text and text-to-speech services.
The following speech-to-speech models are pre-deployed in Live Hub:
-
gpt-realtime
-
gpt-realtime-mini
-
gpt-realtime-1.5
-
gemini-2.5-flash-native-audio
-
gemini-3.1-flash-live
Speech-to-speech models are indicated by
mark in models selector element.
Make sure to select enable voice streaming in Speech and Telephony tab, when working with speech-to-speech models.
For additional information, see Speech-to-speech models.