278 Downloads Updated yesterday
😈 Uncensored Gemma 4 26B (A4B MoE) for Ollama. 👁️ Native Vision, 🛠️ Tool Calling, GGUF, Rust, Linux, Windows AI Coding Agents & Multimodal Local AI Development. 🚀
ollama run jikepjikep_16HEX/gemma-4-26b-nightshift-heretic-uncensored-a4b-moe-q4
Applications
ollama launch claude --model jikepjikep_16HEX/gemma-4-26b-nightshift-heretic-uncensored-a4b-moe-q4
ollama launch codex-app --model jikepjikep_16HEX/gemma-4-26b-nightshift-heretic-uncensored-a4b-moe-q4
ollama launch openclaw --model jikepjikep_16HEX/gemma-4-26b-nightshift-heretic-uncensored-a4b-moe-q4
ollama launch hermes --model jikepjikep_16HEX/gemma-4-26b-nightshift-heretic-uncensored-a4b-moe-q4
ollama launch codex --model jikepjikep_16HEX/gemma-4-26b-nightshift-heretic-uncensored-a4b-moe-q4
ollama launch opencode --model jikepjikep_16HEX/gemma-4-26b-nightshift-heretic-uncensored-a4b-moe-q4
Models
View all →Readme
👁️🗨️ Gemma-4-26B-A4B Nightshift Heretic Uncensored
An uncensored Gemma 4 26B A4B Mixture-of-Experts (MoE) model for Ollama, engineered for multimodal AI, vision understanding, AI Coding Agents, Rust systems programming, Linux development, native Tool Calling, Function Calling, and 256K long-context reasoning.
Built for developers who need fast local inference, advanced image understanding, deterministic code generation, and practical engineering workflows on consumer hardware.
🚀 Overview
Nightshift Heretic transforms Google’s Gemma 4 MoE architecture into a highly optimized local AI model for software engineering, multimodal reasoning, and autonomous development.
Using a 128-expert Mixture-of-Experts architecture, the model activates only the experts required for each token, providing excellent efficiency while maintaining strong reasoning, coding, and vision capabilities.
Optimized for Ollama, GGUF, Rust, Linux, CLI development, AI Coding Agents, MCP workflows, and offline AI environments.
⚡ Key Features
- 😈 Uncensored reasoning with minimal refusal behavior
- 👁️ Native multimodal Vision support (images, screenshots, diagrams, documents)
- ⚡ Efficient A4B Mixture-of-Experts architecture
- 🛠 Native Tool Calling
- 🤖 Function Calling support
- 🦀 Optimized for Rust systems programming
- 🐧 Linux, Ubuntu and CLI development
- 🔧 GGUF optimized for Ollama
- 📚 Ultra-long 256K context window
- 🚀 Deterministic code generation
- 📄 OCR-ready document understanding
- 📊 Chart and diagram interpretation
- 🖥 Optimized for local workstation inference
👁️ Vision Capabilities
Designed for modern multimodal workflows including:
- Screenshot analysis
- Technical documentation
- OCR and document extraction
- UI inspection
- Code screenshot understanding
- Architecture diagrams
- Flowcharts
- Technical drawings
- Charts and graphs
- Image-grounded reasoning
- Vision-guided AI agents
💻 Optimized For
- Ollama
- Claude Code
- OpenClaw
- Codex App
- OpenCode
- Hermes
- MCP-compatible workflows
- AI Coding Agents
- Local AI
- Offline AI
- Software Engineering
- Systems Programming
- Terminal Automation
- Vision-based AI workflows
🦀 Programming Languages & Technologies
Primary languages:
- Rust
- C
- C++
- Zig
- Python
- Go
- Bash
- PowerShell
- JavaScript
- TypeScript
Frequently used technologies:
- Cargo
- Tokio
- Axum
- Docker
- Git
- CMake
- LLVM
- SQL
- Linux APIs
Excellent performance for:
- async programming
- memory safety
- unsafe Rust
- zero-copy pipelines
- multithreading
- backend services
- networking
- embedded systems
- reverse engineering
- performance optimization
⚙️ Native Tool & Vision Calling
Supports modern AI agent workflows with native Tool Calling, Function Calling, and multimodal reasoning.
Ideal backend for:
- Claude Code
- OpenClaw
- Vision-enabled automation
- MCP tool ecosystems
- Local AI assistants
- Autonomous coding pipelines
📊 Benchmarks
| Benchmark | Score |
|---|---|
| GPQA Diamond | 79.2 |
| IFBench | 72.4 |
| AA-LCR | 55.7 |
| Active Parameters | 3.8B / 26B |
⚙️ Configuration
| Parameter | Value |
|---|---|
| Architecture | Gemma 4 26B A4B MoE |
| Experts | 128 |
| Active Parameters | 3.8B |
| Format | GGUF |
| Runtime | Ollama |
| Context Window | 256K |
| Modality | Native Text + Vision |
| Temperature | 0.2 |
| Tool Calling | Native |
| Function Calling | Supported |
🎯 Best Use Cases
- 👁️ Multimodal AI Coding Assistant
- 🤖 AI Coding Agents
- 🦀 Rust Development
- 🐧 Linux Development
- 📷 Image & Screenshot Analysis
- 📄 OCR & Document Processing
- 📊 Diagram Interpretation
- 🔍 Reverse Engineering
- ⚙️ Backend Development
- 🌐 Network Services
- 💾 Embedded Systems
- 🚀 Local AI
- 🔒 Offline AI
- 🧠 Long-Context Reasoning
- 🛠 Technical Automation
🔑 Keywords
Ollama, Gemma 4, Gemma-4-26B, GGUF, A4B, Mixture of Experts, MoE, Vision AI, Multimodal AI, Local AI, Offline AI, AI Coding Assistant, AI Coding Agents, Rust, Linux, Ubuntu, Tool Calling, Function Calling, MCP, OCR, Screenshot Analysis, Document AI, Software Engineering, Systems Programming, CLI Development, Backend Development.
🔬 Design Philosophy
Nightshift Heretic prioritizes objective technical reasoning, deterministic software engineering, and efficient multimodal workflows.
The model is designed to produce accurate source code, analyze technical images and documents, understand long-context projects, and support modern AI-assisted development without unnecessary conversational overhead.
Built for developers who expect high-performance local AI, strong vision capabilities, and practical engineering results from Ollama.