Apple

Discord

Google

The Mistral-Nemo-Instruct-2407 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-Nemo-Base-2407. Trained jointly by Mistral AI and NVIDIA, it significantly outperforms existing models smaller or similar in size.

Second in the Lotus series of high-EQ (emotional intelligence) models by FallenMerick, based on Mistral Nemo. They aim at high quality inference of communication styles to predict mood and add realism to characters' personalities.

Group-chat based roleplaying model trained on an entirely human-based dataset, based on forum/internet group roleplaying styles.

A roleplaying model that is good at following context, realistically portraying characters, and responding creatively. Performs especially well for its size.

The one and only Space Whale, remastered and requantized in full 32-bit precision. Should be a dramatic improvement over the classic Psyonic Cetacean 20B in terms of creativity and quality of inference.

Finetuned from Llama 3.1, with better instruction following and character personification. Performs well in a variety of languages.

Uncensored Llama2 model that was designed for roleplaying and storytelling and excels at both. The model performs well at other tasks and responds well to changes in the prompt.

A direct replacement/successor to Euryale v2.2, entirely trained on top of Llama 3.3 Instruct.

First in a series of models designed to replicate the prose quality of the Claude 3 models, specifically Sonnet and Opus. This model is fine-tuned on top of Qwen-2 72B Instruct.

Version 3.3 of the popular Stheno series. A model made for 1-on-1 roleplay, but also able to handle scenarios, RPGs and storywriting. Uncensored during actual roleplay scenarios.

Research release of a 104 billion parameter highly performant generative model optimized for reasoning, summarization, and question answering. Command-R supports multilingual generation in 10 languages and has RAG capabilities.

A lightweight model that thinks before responding. Fast, smart, and great for logic-based tasks that do not require deep domain knowledge. The raw thinking traces are accessible.

Llama3-based RP and Storytelling model by Hastagaras, with DPO and LORAs applied to reduce cheerfulness and emoji.

Model Statuses

Mistral Nemo Instruct 2407 12B

MN Violet Lotus 12B

MN BackyardAI Party V1 12B

Fimbulvetr v2 11B

Psyonic Cetacean V1 Ultra Quality 20B

Llama 3.3 Instruct 70B

Midnight Rose 70B

L3.3 Euryale V2.3 70B

Magnum V1 72B

L3 Stheno V3.3 32K 8B

C4ai Command R Plus 104B

Grok 3 Mini

Jamet L3 MK.V Blackroot 8B