Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
Nous Research has launched Hermes 2 Pro, a new model based on Meta's Llama-3 8B architecture, which is now available on HuggingFace. The Hermes 2 Pro features enhanced capabilities such as Function Calling and Structured Output, utilizing dedicated tokens for tool call parsing tags. This model not only matches but also surpasses the performance of Meta's own Llama-3 8B model, achieving this within two weeks of Llama-3's release. The Hermes 2 Pro, an instruct model, is optimized for specific instruction-based responses, contrasting with the probabilistic outputs of other models.
Mind blowing 🤯 function calling by the new `Hermes 2 on Llama-3` by @Teknium1 @intrstllrninja running on @ollama The last question asks it to do 3 different function calls and write an article. Just check that quality. code: https://t.co/QLscvoY45i https://t.co/xYIiOeLZea
Hermes 2 Pro on Llama-3 8B just released. 🔥 @NousResearch released their first Llama-3 fine-tune, outperforming Meta's Llama 3 8B instruct. 🤯 Hermes Pro is optimized for Function Calling and Structured Output capabilities using a dedicated token for tool call. TL;DR: 🧠 based… https://t.co/bYmGO0YXZx
Nous Research released an instruct model of Metas new Llama-3 model that beats Meta’s own version Llama-3 came out 2 weeks ago so they did it within this time frame. Instruct LLMs offer more specific instruction based answers vs probabilistic LLMs Ridiculously good team 🐐👏 https://t.co/H4UTH5gWx3
