Indexnusrefather/Super-Slop-Machina-Roleplay-1.2b-ERP-Tolerant-V2

TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.2BQuant:BF16Context Size:32kPublished:Jul 21, 2026License:lfm1.0Architecture:Transformer0.0K Featherless Exclusive Cold

Indexnusrefather/Super-Slop-Machina-Roleplay-1.2b-ERP-Tolerant-V2 is a 1.2 billion parameter language model with a 32768-token context length, fine-tuned by Indexnusrefather for enhanced roleplay capabilities. This second iteration builds upon its predecessor with extended training data and an aggressive methodology to improve writing style. It is specifically optimized for generating high-quality, coherent roleplay interactions.

Loading preview...

Overview

Indexnusrefather/Super-Slop-Machina-Roleplay-1.2b-ERP-Tolerant-V2 is the second iteration of a 1.2 billion parameter model specifically fine-tuned for roleplay. It features a 32768-token context length and aims to provide improved roleplaying performance compared to its previous version.

Key Improvements from V1

  • Extended Training Data: The model's training was significantly extended from approximately 35 million to 52 million tokens, providing a broader base for learning.
  • Aggressive Methodology: An improved, more aggressive training methodology was employed to specifically enhance the model's ability to learn and adapt to new writing styles, directly translating to better roleplay quality.

Quantization Recommendations

The model offers various quantization options, now improved with imatrix for better quality preservation. Recommendations for use include:

  • BF16: Highest quality with the fewest logical mistakes.
  • Q8_0: High quality, near-lossless, with only slightly more mistakes.
  • Q6_K: A good balance if Q8_0 is too demanding, with minor detail loss.
  • Q5_K_M: Suitable for hardware constraints, though degradation becomes noticeable.
  • Q4_K_M: Generally not recommended due to clear degradation.
  • IQ4_XS: The lightest option, using imatrix for coherence, but with potential for significant degradation.

Future Development

The developer plans to further extend training to approximately 70 million tokens for a potential V3 release, indicating ongoing commitment to enhancing the model's roleplay capabilities.

Finetunes
1 models