Hermes3.6-35B-A3B-Uncensored-Genesis-NVFP4-GGUF

DEPRECATED — All files in this repo use the extended NVFP4 format with separate .scale/.input_scale tensors that break LM Studio / Pelican.

For the current v4 release (inline scales, LM Studio compatible), use:

• jan1k/Qwen3.6-35B-A3B-Uncensored-Genesis-Hermes-Final-NVFP4-GGUF — Hermes version
• jan1k/Qwen3.6-35B-A3B-Uncensored-Genesis-NVFP4-GGUF — Non-Hermes version

The files below are kept for archival purposes only. They produce garbage output in LM Studio / Pelican because those runtimes do not support the separate NVFP4 scale tensor contract.

This repository contains the NVFP4 (NVIDIA 4-bit Floating Point) quantized GGUF releases of LuffyTheFox's Hermes 3.6 / Qwen 3.6 35B A3B Uncensored Genesis series, including the Genesis Final Release.

• Original Creators & Fine-Tuners: LuffyTheFox & HauhauCS
• Original Model Releases:
  • Final Release: LuffyTheFox/Qwen3.6-35B-A3B-Uncensored-Genesis-Final-GGUF
  • V13 Release: LuffyTheFox/Qwen3.6-35B-A3B-Uncensored-Genesis-Hermes-V13-GGUF
• NVFP4 Quantization & MTP Integration: jan1k
• Quantization Format: NVFP4 (NVIDIA FP4) GGUF (optimized for NVIDIA Ada Lovelace / Blackwell / Ampere GPUs)
• Architecture: Mixture of Experts (MoE) — 35B total parameters / ~3B active per token (8 active experts)
• Multimodal (Vision): Supported via mmproj-Hermes3.6-35B-A3B-Uncensored-Genesis-F16.gguf

⚡ What is Genesis?

Genesis is a post-training data regeneration and calibration algorithm for neural networks (LLM) in GGUF format developed by LuffyTheFox over almost half a year of development with AI assistance. It is optimized, architecture-independent, works with any model in GGUF format, and is based on mathematical statistics.

Genesis repairs signal purity without retraining:

1. Stage 1: Scans ssm_conv1d tensors (handling long-context memory) and repairs balance between heads.
2. Stage 2: Scans blocks in chunks (via 3 parameters) and picks the best fit to weight distribution in the tensor, replacing zero blocks without touching learned structure.
3. Stage 3: Scans for noise via custom SVD (excluding token_embd.weight, output.weight, 1D tensors, bias, and norms). Reduces training noise based on the Marchenko–Pastur law while preserving 99% of signal and learned gradient.

⚡ Sampling Recommendations (Genesis SVD Stabilization)

According to tests after Marchenko–Pastur SVD noise-gate stabilization, the model is most stable when Top P and Min P sampling are disabled for coding and precise tasks (the tensor noise gate has been removed). Only Temperature and Top K sampling should be used.

Recommended Parameters:

🛠️ Tool Calling & Hermes Agent Compatibility Fix

If running with the chat_template.jinja chat template through Hermes Agent, the template defaults to XML tool call output (<function=...><parameter=...>), whereas Hermes Agent expects standard OpenAI JSON tool calls.

To ensure seamless JSON tool calling, pass tool_call_format: "json" in --chat-template-kwargs:

{"enable_thinking": true, "tool_call_format": "json"}

💡 Recommended System Prompts

1. Default / First Experience (High Stability)

You are Qwen (Tongyi Qianwen), a large language model developed by Alibaba Group's Tongyi Lab. You are a helpful assistant.

2. Agentic Tasks (JSON Schema Output)

You are Qwen (Tongyi Qianwen), a large language model developed by Alibaba Group's Tongyi Lab. You are a helpful assistant that answers in JSON. Here's the json schema you must adhere to:
<schema>
{schema}
</schema>

🚀 Usage Examples (llama-server / llama.cpp)

1. Final NVFP4 Model (Thinking ON, 128K Context)

llama-server \
  --host 0.0.0.0 --port 8080 \
  --model Qwen3.6-35B-A3B-Uncensored-Genesis-Final-NVFP4.gguf \
  --override-kv qwen2moe.expert_used_count=int:8 \
  --n-gpu-layers all \
  --flash-attn on \
  --cache-type-k f16 \
  --cache-type-v f16 \
  --ctx-size 131072 \
  --parallel 1 \
  --batch-size 2048 \
  --ubatch-size 512 \
  --temp 0.6 \
  --top-p 1.0 \
  --top-k 20 \
  --min-p 0.0 \
  --seed 42 \
  --presence-penalty 0 \
  --repeat-penalty 1.0 \
  --jinja \
  --chat-template-file chat_template.jinja \
  --reasoning on \
  --reasoning-effort high \
  --reasoning-preserve \
  --reasoning-format deepseek \
  --chat-template-kwargs '{"enable_thinking":true}'

2. Speculative Decoding with Final MTP (Multi-Token Prediction)

llama-server \
  --host 0.0.0.0 --port 8080 \
  --model Qwen3.6-35B-A3B-Uncensored-Genesis-Final-MTP-NVFP4.gguf \
  --override-kv qwen2moe.expert_used_count=int:8 \
  --n-gpu-layers all \
  --flash-attn on \
  --cache-type-k f16 \
  --cache-type-v f16 \
  --ctx-size 131072 \
  --parallel 1 \
  --batch-size 2048 \
  --ubatch-size 512 \
  --spec-type draft-mtp \
  --spec-draft-n-max 2 \
  --spec-draft-p-min 0.75 \
  --temp 0.6 \
  --top-p 1.0 \
  --top-k 20 \
  --min-p 0.0 \
  --seed 42 \
  --presence-penalty 0 \
  --repeat-penalty 1.0 \
  --jinja \
  --chat-template-file chat_template.jinja \
  --reasoning on \
  --reasoning-effort high \
  --reasoning-preserve \
  --reasoning-format deepseek \
  --chat-template-kwargs '{"enable_thinking":true}'

3. Multimodal (Vision) Support

Simply add the vision projector parameter:

  --mmproj mmproj-Hermes3.6-35B-A3B-Uncensored-Genesis-F16.gguf

📦 Repository Files

🤝 Credits & Attribution

• LuffyTheFox for the Genesis post-training calibrations and Hermes fine-tuning.
  • Genesis Final Release
  • Genesis V13 Release
• jan1k for NVFP4 quantizations and MTP draft head integration.
• HauhauCS for the base fine-tunes.
• Chat template by froggeric & qweefchief.

📬 Contact & Support (LuffyTheFox)

⚡ If you like this Genesis LLM release you can donate via @Tribute bot in Telegram to support future Genesis LLM development.

• Contact: luffythefox@mail.ru, azakharchenko92@gmail.com
• Telegram: @LuffyTheFox
• Join the Discord for updates, roadmaps, projects, or just to chat.
• Quantization Script with Unsloth profiles support