CMSManhattan / JiRackUltra_7b

JiRack Ultra 7B (CPU)

A fast and efficient 7B model optimized for CPU inference. The model was refactored with BitNet features and an updated tokenizer that includes new Routing , Media , Vision , Sound , Tool call , and Robotics tags. Built on a Qwen2.5-7B architecture with native ternary (BitNet-style) support and ready-to-run GGUF quantizations.

• JiRack is a cloud-ready model that helps save money on cloud infrastructure. It can be used as an expert model in RAG deployments, with the ONNX JiRack Java server as an alternative.
• Subscription: $1 per month per user (updated license for non-company use).
• Corp Subscription: $3 per month per user (updated license for company use).
• It works without subscription but send message about subscription

Available Variants

Quick Start

Run with Docker

Default CPU (Q4 recommended)

docker run -d \
  --name jirack_ultra_7b \
  -p 7869:7869 \
  --restart unless-stopped \
  cmsmanhattan/jirack-ultra-7b-cpu-q4:latest

Q3

docker run -d \
  --name jirack_ultra_7b \
  -p 7869:7869 \
  --cpus=16 \
  -e THREADS=16 \
  -e THREADS_BATCH=16 \ 
  --restart unless-stopped \
  cmsmanhattan/jirack-ultra-7b-cpu-q3:latest

Q2 (lowest memory)

docker run -d \
  --name jirack_ultra_7b \
  -p 7869:7869 \
  --cpus=16 \
  -e THREADS=16 \
  -e THREADS_BATCH=16 \ 
  --restart unless-stopped \
  cmsmanhattan/jirack-ultra-7b-cpu-q2:latest

Full precision

docker run -d \
  --name jirack_ultra_7b \
  -p 7869:7869 \
  --cpus=16 \
  -e THREADS=16 \
  -e THREADS_BATCH=16 \ 
  --restart unless-stopped \
  cmsmanhattan/jirack-ultra-7b-cpu:latest

Multi CPU

docker run -d \
  --name jirack_ultra_7b \
  -p 7869:7869 \
  --cpus=16 \
  -e THREADS=16 \
  -e THREADS_BATCH=16 \ 
  --restart unless-stopped \
  --memory=16g \
  --cpus=8 \
  cmsmanhattan/jirack-ultra-7b-cpu-q4:latest

Docker Compose Example

services:
  jirack:
    image: cmsmanhattan/jirack-ultra-7b-cpu-q4:latest
    container_name: jirack_ultra_7b
    ports:
      - "7869:7869"
    volumes:
      - .:/app
      - ./web:/app/web
    environment:
      - MAX_TOKENS=2048
      - TEMPERATURE=0.7
      - TOP_P=0.9
      - DEFAULT_STREAM=False
      - INTRA_THREADS=4
      - USE_ENV_ALLOCATOR=1
      - THREADS=16 
      - THREADS_BATCH=16 
    deploy:
      resources:
        limits:
          memory: 16g

Access the UI

Once the container is running, open your browser and navigate to:

http://localhost:7869

This opens the JiRack UI — a clean web interface.

Changing the Port

The listening port can be easily modified directly from the Settings panel within the JiRack UI.

Licensing

• The JiRack Ultra 7B model is provided under a commercial license ($12 per user per year).
• All JiRack UI clients are provided under a commercial license.
• However, the UI clients can be used for free when running together with the official JiRack Docker containers, as long as they are not redistributed separately.

For commercial licensing, cluster deployment, or enterprise use of JiRack models, please contact us.

• JiRack MS Windows 11 Desktop Client (with Ollama API): https://huggingface.co/kgrabko/JiRackTernary_1b/resolve/main/jirack-chat.zip
• Live email chat with the model: [email protected]

Hardware Recommendations

Recommended Hardware for JiRack Ultra 7B (single Docker container)

Important Memory Notes

Even though the quantized 7B models are small, we recommend the following for best experience:

• Q4_K_M: 8–12 GB system RAM minimum
• Q3_K_M / Q2_K: 6–10 GB system RAM
• Full precision: 16 GB+ system RAM recommended

Reasons for extra headroom:

• KV-cache consumption during generation
• Runtime overhead and temporary buffers
• System stability and avoiding out-of-memory errors
• Room for larger context windows

Minimum recommended (Q4): 12 GB system RAM
Ideal: 16–24 GB system RAM

I added the default model in full precision. This serves as the base for quantization, allowing us to find the optimal balance between model size and performance.

Architecture Notes

• Refactored with BitNet features : Native BitLinear ternary path (b1.58-style) with λ-warmup STE
• Updated tokenizer : Extended with new special tags for Routing , Tool call , and Robotics
• Base: Qwen2.5-7B style (Hidden 3584, 28 layers, GQA 28/4, vocab 152064)
• RoPE θ = 10000, RMSNorm ε = 1e-6
• Ready-to-run GGUF quantizations (Q2_K, Q3_K_M, Q4_K_M)

📧 Contact & Licensing

For joint venture opportunities, hardware integration, or licensing inquiries:

• Email: [email protected]
• Phone: +1 (516) 777-0945
• Location: New York, USA