Hermes 4 expands Nous Research's line of neutrally-aligned and steerable models with a new group of hybrid reasoners. Like all of our models, these are designed to adhere to the user's needs and system prompts, rather than to a company's ethics code. Hermes users will feel an eagerness for roleplaying and creativity in the model. They'll also notice a lack of lecturing and sycophancy. Put simply, Hermes users will experience a more pleasant, humanistic interaction.
Because they are "reasoning" models, they have the ability to "spend" more tokens during inference to think longer about hard problems. Hybrids can toggle between reasoning and standard responses by including or omitting a <think> tag with a request. By being hybrid, Hermes improves performance across a number of benchmarks and maintains efficiency when thinking isn't necessary.
The training dataset for Hermes 4 expands over the dataset for Hermes 3 with 50x more data tokens. Much of the data was synthesized using Atropos, and special care was taken to highlight creative content.
Nous Chat has been revamped to serve these new, powerful models. Nous Chat gives users more hands-on control of the model and includes features like completions mode, etc.
Chat With HermesRead The Technical ReportDownload the Models
Hermes 4 is available to try with custom system prompts, chat formats, and memory on our newly rebuilt Nous Chat web application. Nous Chat is built for power users and casual conversations alike. Should one wish for more fine-tuned control of the model, the settings panel exposes a breadth of parameters to customize outputs.
Completions Mode gives users a unique experience with the model by providing access to a wider variety of chat templates, which the H4 models adapt favorably to. The chat-window layout allows users to easily switch between concurrent conversations.
When coupled with custom workspaces, users can quickly navigate environments with different models and settings. The experience will be more similar to conversations with a phonebook of various experts and personalities, rather than a dialogue with a monolithic character in different contexts.
In a step away from the anthropomorphism affecting user-assistant interactions, we bring you the orb. The classic depiction of AI in both cinema and games is very dear to our hearts and our aesthetic--simultaneously, it is also a functionally effective interface. This orb on the right represents the continuous memory system users will find in Nous Chat.
Users can store memories in a workspace orb, then switch between different prompt templates and models with a consistent knowledge graph.
This collection of memories and relationships is unique to each user's individual workspaces, so one can easily populate separate memory orbs for different functions. We are working to add editable memories for users through the orb as we aim for a experience that becomes both more modular and easier to use as we iterate on Nous Chat.