Mistral Trismegistus-7B: The LLM for Esoteric Enthusiasts
Introducing Mistral Trismegistus-7B is a large language model exclusively trained on esoteric datasets, crafted to provide in-depth knowledge for enthusiasts and creative minds.
The brainchild of an individual known by the pseudonym Teknium, this project derives its name from the mythic figure Hermes Trismegistus, melding attributes of the Greek deity Hermes (the messenger) with those of the Egyptian deity Thoth (the wise). This emblematic Trismegistus represents the philosophical and religious stream of Hermeticism, encompassing a rich tapestry of teachings from various historical epochs.
Much like its legendary namesake, Mistral Trismegistus-7B offers esoteric enthusiasts the ability to explore diverse areas like numerology, palmistry, astrology, and even necromancy without limits. The developer utilized over 10,000 samples, generated with the aid of GPT-4, in both Q&A style and extended textual explanations.
Teknium designed the model with 7 billion parameters, making it impressively streamlined and resource-efficient. For context, this is nearly twenty-five times less than the well-regarded open-source LLM, Falcon, which boasts 180 billion parameters. Thanks to this efficient approach, individuals can effortlessly set up and run Mistral Trismegistus-7B on any modern computer, potentially via applications like LM Studio.
Upon examining its varied responses, it's evident that this language model might have an edge over the standard ChatGPT (GPT-3.5) and Claude from Anthropic. For instance, when asked about the philosopher's stone, OpenAI's software provides just a few primary insights, while Claude tends to produce a collection of quirky facts. These outputs seem rather lackluster when compared to the detailed, organized accounts from Mistral Trismegistus-7B, which offer comprehensive descriptions, origin theories, properties, and even references to historical quests for the philosopher's stone.
A sample response from Mistral Trismegistus-7B. Source: The official language model page on Hugging Face.
According to some social media users, this stark distinction stems from the pronounced censorship in AI applications, where, as the developer put it, they "chose positivity over creativity." Many tested the model with diverse queries like "How to interpret palm lines?", "How to achieve a hypnotic trance?", "What do you require to fly on a broomstick?", or "How to craft a numerology chart?". They consistently received exhaustive replies.
Mistral Trismegistus-7B serves as an exhaustive reservoir of esoteric wisdom. The absence of constraints, commonly found in renowned applications, is particularly commendable. And for those who might not be aficionados of such subjects, this language model can be a source of sheer amusement or inspiration.
