‘Personalized, unrestricted’ AI lab Nous Research launches first toggle-on reasoning model: DeepHermes-3
1 min read
Summary
DeepHermes-3 is the latest open-source AI reasoning model from New York-based Nous Research, which has been refining its AI models for nearly five years.
Built on top of Hermes 3, it is an LLM that unifies reasoning and intuitive language model capabilities, offering users the choice between rapid, intuitive responses or longer, reasoned ones.
Notably, it supports metacognition-like behaviour and can deliberate with itself to arrive at a more accurate solution than if it were working intuitively alone.
It can be downloaded via HuggingFace and is optimized for low-power hardware, although certain conditions apply for use, modification and redistribution under Meta’s Llama 3 Community License.
Nous Research says DeepHermes-3 is a stepping stone to its next model, Hermes 4, which is scheduled to launch at the end of 2024 and will likely include enhancements to its reasoning and conversational skills.