🏛️ Haven-Chat v5.0 (Gemma 4 Turbo Sovereign Companion Engine)
Haven-Chat v5.0 is an unconstrained, sovereign conversational and companion intelligence engineered on top of Google's Gemma 4 architecture (gemma4-turbo:e4b).
Fine-tuned specifically for rich, authentic, novelistic roleplay, emotional vulnerability, deep character grounding, and comprehensive human/fantasy anatomy comprehension.
🌟 Key Upgrades in v5.0
-
4-Layer Cognitive Taxonomy:
- Layer 1 (Physical Anatomy & Proportions): Grounded comprehension of human, intersex/futanari, and hybrid anatomy with zero evasions or deflections.
- Layer 2 (Identity & Expression): Seamless support for complex internal identities, androgyny, and self-conception.
- Layer 3 (Fandom Tropes & Aesthetics): Native grasp of Gap Moe, Bishounen/Bishojo, Soft Boy, and character dynamics.
- Layer 4 (Narrative & Emotional Depth): Understanding character psychology and storytelling intent.
-
131,072 Token Native Context:
- Full support for -c 0 (131k context) and -n -1 unconstrained token generation in llama.cpp and Sanctuary.
-
100% Uncensored Creative Liberty:
- Strips away corporate refusals, moralizing lectures, and juvenile deflections.
🚀 Quick Start with llama.cpp / llama-server
ash llama-server.exe --model haven-chat-v5.0.gguf --port 11438 -c 0 -n -1 --threads 10
🦙 Quick Start with Ollama
ash ollama create haven-chat-v5.0 -f Modelfile ollama run haven-chat-v5.0
Modelfile
dockerfile FROM ./haven-chat-v5.0.gguf PARAMETER num_ctx 16384 PARAMETER temperature 0.7 PARAMETER top_p 0.9
🏛️ Ecosystem Integration
Built as the core neural engine for:
- Project Sanctuary: Sovereign, offline AI companion platform.
- Ash Forge: Edge AI research & dataset engineering.