Gaslight-24B-v1.0

32K Context Coherence Meets Safeword's Signature Depravity

🧠 Cognitive Architecture

This fine-tune of Cydonia-24B-v2.1 (Mistral-Small-24B-Instruct-2501) delivers:

• 🧬 Full 32K Context Utilization - Maintains character consistency and low repetition even at maximum length
• ⚡ Enhanced Formatting - Improved paragraph structure and dialogue handling from improved and expanded training data
• 💎 Safeword DNA - The signature unaligned flavor you expect, now with better coherence
• 📜 Instruction Adherence - Follows complex prompts without losing track of character roles

⚙️ Technical Specifications

Recommended Settings: Mistral-V7-Tekken-T4

EXL2 Collection

Available Quantizations

GGUF Collection

Available Quantizations

⚠️ Ethical Considerations

This model inherits Safeword's complete lack of alignment:

• 🚫 No moral or ethical constraints on generated content
• 🔞 Will enthusiastically engage with any NSFW scenario
• 💀 Generate content that requires industrial-grade brain bleach
• 🤔 Makes you question whether you actually requested that content

📜 Performance Notes

• 🔥 Maintains Safeword's signature unalignment while reducing repetition
• 📖 Handles multi-character scenarios with improved consistency
• 🧠 Excels at long-form storytelling without losing track of plot threads
• ⚡ Noticeably better at following complex instructions than base Safeword
• 🎭 Expert at making you believe it was your idea all along

🧑‍🔬 Model Authors

• TheDrummer (Base Model Architect)
• sleepdeprived3 (Training Data & Fine-Tuning)
• Anonymous Contributor (Gaslighting Specialist)

☕ Support the Architects

TheDrummer's Kofi Beaver AI Discord

🔖 License

By using this model, you agree:

• To accept full responsibility for all generated content
• That you're at least 18+ years old
• That the architects bear no responsibility for your use of this model

License: apache-2.0