Hello all, what is the difference between inference and plugins. Does inference will be fast compared to plugins?
suggest me the best and fast LLM in term of latency.
what is the difference between inference and plugins.
The best reference to describe LiveKit inference is LiveKit Inference | LiveKit Documentation, it provides provides access to many of the best models and providers for voice agents, including models from OpenAI, Google, AssemblyAI, Deepgram, Cartesia, ElevenLabs, and more. LiveKit Inference is included in LiveKit Cloud, and does not require any additional plugins.
Plugins require you to manage your own keys and have separate accounts set up with each plugin provider, so it can be more overhead and a higher barrier to entry compared with LK inference.
Plugins are arguably more flexible than LiveKit inference however, since more providers are available and not all features in Plugins are available in LK inference yet (such as custom voices, though we are working to close those gaps)
Does inference will be fast compared to plugins?
In our internal testing, there is no difference between the two.
suggest me the best and fast LLM in term of latency.
You could make the case that those two things are opposite 
the faster an LLM is at reasoning the less capable it is likely to be. Most developers try to find the ‘sweet spot’ between the two (gpt-4.1-mini is popular) - if you go to our homepage livekit.io and click ‘talk to agent’ you can experiment with several agents, some of whom use different models, rather than set it all up yourself.