Venice AI is a service that offers a simple interface to open-source AIs. Venice has now enabled Meta’s brand new open-sourced frontier model Llama 3.1 405B. It’s the largest published to date, with 405 billion parameters.
However, note that 405B is not connected to the internet. It’s best to use the web-enabled version of Nous Theta for questions that require real-time information. 405B will be slower relative to smaller models. This model is still somewhat censored, although Venice expects to be able to introduce an uncensored version soon.
Meta’s release of this LLM demonstrates that open-source models can exceed the capabilities and performance of many centralized, closed-source models.
More About Meta’s Newest LLM
“Llama 3.1 405B” is the latest and most advanced open-source large language model (LLM) from Meta.
Here are some key features:
1. Size Of Model
- Llama 3.1 has 405 billion parameters. The more parameters, the smarter the model.
2. Extensive Training:
- This model was trained on 15.6 trillion tokens (think of tokens as pieces of text) using 16,000 powerful GPUs. This extensive training helps it understand language better and respond more accurately.
3. Long Context:
- It can handle a lot of information at once, with a context length of 128,000 tokens. This means it can remember and process longer conversations or texts without losing track.
4. Open Source:
Llama 3.1 is available for everyone to use and modify. You can download it from platforms like Hugging Face.