Free and Open Source: Since it is an open-source model, it is completely free to use, whether running it locally or through APIs provided by some service providers.

What truly distinguishes this model, in my opinion, is its exceptional ability to follow complex instructions and adhere precisely to the required format. The Nous Research team has done an excellent job in fine-tuning using the ChatML format, which allows us to impose strict and highly controlled System Prompts.

UI / UX (User Interface and User Experience) - From an independent perspective:

The user experience with Hermes 3 varies depending on the method of operation, but generally, I find it flexible and straightforward, with some notes:

1. Operation via Command Line Interface (CLI):

Experience: Since I run the model locally on my device using llama.cpp, the primary interaction is through the command line. The interface is simple and free of complications, requiring only to write a command like:

./llama-cli -m hermes-3-70b.Q4_K_M.gguf -p "your prompt here" Review collected by and hosted on G2.com.

The biggest weakness is the feature of tool/function calling, as it is not reliable enough for use in real client projects.

As a freelancer, I try to automate as many tasks as possible to reduce the time I spend on each project and increase my profits. I tried using Hermes 3 to create an intelligent agent that connects to an SQL database to query client data, but the experience was disappointing.

The first problem: When I ask it to execute two queries at the same time (such as: "Find the number of pending orders and the total sales for this week"), it often fails to execute the second query, then fabricates an incorrect number. The failure rate reached about 60% of attempts, which makes the feature unsuitable for reliance in a production environment. Review collected by and hosted on G2.com.