Octopus V2 : on-device language model with 2 billion parameters
Are you still confused ? Not only Now this is a light model that can be run in your mobile device but with comparatively good accuracy and latency than someone which I will mention below.
Introduction:
Octopus v2 is developed by Nexa AI .
Why this ?
There is a great concern regarding our security and details provided to distant server. What if that operates in our mobile device. Wow. Sure will use it. But current models have great latency and accuracy issue where this comes to pull from the hole.
This is a great standard set of accuracy and speed in mobile device. It has surpassed GPT-4 in both accuracy and latency.
Successor models
Along with this there are v3, v4 models are released. I have not written much of the content but do check this .
Limitation: limitation exists so be patience.
References:
1.https://huggingface.co/NexaAIDev/Octopus-v2
2.https://tryopen.ai/octopus-v2-fast-on-device-ai-for-android/
Diving deeper. ?
Check this out
3. https://arxiv.org/abs/2404.01744