RicardoEstep/RPBizkitRemiX-v3-12B

TEXT GENERATIONPricing:Input $1.2 / Output $4.8Concurrent Unit Cost:1Model Size:12.2BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 26, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

RicardoEstep/RPBizkitRemiX-v3-12B is a 12.2 billion parameter experimental Mistral Nemo mix model created by RicardoEstep, utilizing a DARE TIES merge of several RP Bizkit models. This model is designed to be a cleaner, uncensored remix, offering improved stability and independence from previous versions. It features a clean tokenizer and embedding sizes (131072), and is intended to support a 128K context size, making it suitable for creative roleplay and extended conversational tasks.

Loading preview...

RicardoEstep/RPBizkitRemiX-v3-12B: Experimental Mistral Nemo Mix

This model, developed by RicardoEstep, is an experimental 12.2 billion parameter language model based on a "Mistral Nemo" mix. It leverages a DARE TIES merge using Mergekit to combine several specialized models, including TheDrummer/UnslopNemo-12B-v4.1, RicardoEstep/RPBizkit-v6-12B, and RicardoEstep/RPBizkit-v9-12B.

Key Characteristics & Improvements

  • "Cleaner" RemiX: Addresses and resolves a "noisy" hard "assistant image" issue found in previous versions, even at low weights.
  • Uncensored Output: The base model has been changed to ensure greater independence, resulting in a completely uncensored model.
  • Extended Context: Features clean tokenizer and embedding sizes (131072) and is designed to support a 128K context size.
  • Recommended Usage: The model is pre-configured to not use any chat templates, with "Alpaca (with "RAW" inputs)" being the recommended chat template. Lower temperatures (0.86-0.92) are suggested for creative roleplay.

Use Cases

This model is particularly well-suited for:

  • Creative Roleplay (RP): Optimized for generating creative and engaging roleplay scenarios with recommended lower temperatures.
  • Extended Conversational Tasks: The large context window (128K) allows for maintaining coherence over very long interactions.
  • Uncensored Content Generation: Designed to provide unrestricted outputs, making it suitable for applications requiring full creative freedom.

Finetunes
2 models