Indexnusrefather/Super-Slop-Machina-Roleplay-1.2b-V4-ERP-Tolerant-it
Indexnusrefather/Super-Slop-Machina-Roleplay-1.2b-V4-ERP-Tolerant-it is a 1.2 billion parameter language model, post-trained for enhanced roleplay capabilities. This version, adapted for longer context lengths of 32768 tokens, consumes higher quality data to significantly improve roleplay interactions. It is specifically designed to excel in detailed and immersive roleplaying scenarios, offering better performance than its preview version.
Loading preview...
Overview
Indexnusrefather's Super-Slop-Machina-Roleplay-1.2b-V4-ERP-Tolerant-it is a 1.2 billion parameter model that has undergone significant post-training (SFT) to specialize in roleplay. This iteration, building upon its preview version, focuses on delivering a superior roleplaying experience.
Key Improvements & Capabilities
- Enhanced Roleplay Quality: The model has been specifically adapted and fine-tuned with higher quality data, resulting in substantially improved roleplay interactions compared to its predecessors.
- Extended Context Length: It is designed to handle longer contexts, supporting up to 32768 tokens, which is crucial for maintaining coherence and depth in extended roleplay scenarios.
- Optimized Data Consumption: The training process involved consuming higher quality data, directly contributing to its refined performance in generating nuanced and engaging roleplay content.
Recommended Quantizations
For optimal performance, the creator recommends specific quantization levels based on hardware capabilities:
- BF16: Offers the highest quality with the fewest logical mistakes.
- Q8_0: Provides high quality, with only slightly more mistakes, considered near lossless.
- Q6_K: A viable option if Q8_0 is too demanding, though minor detail loss may occur.
- Q5_K_M: Recommended only for severely limited hardware, where degradation becomes noticeable.
Future Developments
The developer is actively working on models under 1 billion parameters and experimenting with Qwen 3.5 9b, with promising results anticipated for future releases.