Llama 3.2 and SmolLM2 Lead AI Advancements with 1B and 3B Models, Cybersecurity Applications, and Zero-Shot Voice Cloning

Recent advancements in AI models have been highlighted, particularly with the release of Llama 3.2, which includes new vision language models and compact versions with 1 billion and 3 billion parameters. These models are pre-trained and aligned, with drop-in vision capabilities, and are available for use through Meta AI and Hugging Face. Additionally, a local real-time voice mode based on Llama 3 has been introduced, which is open-source and free. Another model, 'llama-3.1-whiterabbitneo-2-8b', has been released for cybersecurity applications as a public preview. This model can be installed and run using Local AI. Furthermore, Hugging Face has launched SmolLM2, a compact AI model that outperforms larger LLaMa models while utilizing only 1/8th of their parameters. This model is available on Bakery by Bagel Network. Lastly, new Smol TTS models have been introduced, featuring zero-shot voice cloning based on LLaMa architecture, which operates on-device with llama.cpp.