Hanzo Notebooks: Managed Jupyter with GPU Access
Introducing Hanzo Notebooks — managed Jupyter workspaces with on-demand GPU access and pre-installed ML frameworks.
Setting up a machine learning development environment is a two-day project: install CUDA drivers, resolve version conflicts between PyTorch and CUDA, configure Jupyter, set up remote access, attach storage for datasets. And then do it again when you need a different GPU.
Today we are launching Hanzo Notebooks: managed Jupyter workspaces that start in 30 seconds with the GPU you need.
How It Works
Instant Workspaces
Click a button, get a Jupyter environment:
- Pre-installed: PyTorch, TensorFlow, JAX, Hugging Face Transformers
- CUDA 12.x configured and ready
- Python 3.11+ with common ML libraries
- Terminal access for custom installations
GPU Selection
Choose the GPU that matches your workload:
| GPU | VRAM | Use Case | Price |
|---|---|---|---|
| T4 | 16GB | Inference, small fine-tunes | $0.50/hr |
| A10G | 24GB | Medium training, inference | $1.00/hr |
| A100 40GB | 40GB | Large model training | $3.00/hr |
| A100 80GB | 80GB | Multi-billion parameter models | $5.00/hr |
| H100 | 80GB | Frontier model training | $8.00/hr |
Switch GPUs without recreating your environment. Your files persist across GPU changes.
Persistent Storage
- Home directory persists across sessions
- Shared team storage for datasets
- S3-compatible access to Hanzo Storage
- Git integration for version control
Collaboration
- Share notebooks with team members
- Real-time collaborative editing
- Comment and review workflows
- Version history for every notebook
Integration with Hanzo ML Platform
Notebooks integrate natively with the broader Hanzo ML infrastructure:
- Submit training jobs to Hanzo Training from a notebook
- Access models in Hanzo Registry
- Query features from Hanzo Feature Store
- Log experiments to Hanzo Console
Pricing
- Free: 20 GPU-hours/month (T4)
- Pro: $49/month — 100 GPU-hours, all GPU types
- Team: $99/month per seat — shared storage, collaboration
- Enterprise: Reserved GPU capacity, custom images
Hanzo AI is Techstars '17.
Read more
accel: One Dispatch Fabric for GPU, CUDA, Metal, and FPGA
LP-138 lands. accel is a capability-negotiated dispatch layer so accelerated crypto, ZK, and FHE are one codebase across runtimes.
Hanzo ML Platform: End-to-End Machine Learning Infrastructure
Announcing the unified Hanzo ML Platform — training, serving, pipelines, feature store, model registry, and evaluation in one integrated system.
Introducing Enso — one API, routed to the best frontier intelligence
Enso is Hanzo's learned router: no single frontier model leads every benchmark, so Enso routes each request to the model most likely to win it — and learns from your feedback as it goes. Microsecond routing on a CPU, a transparent per-request meter, and a 1% fee that pays for itself. Here is how it works, the measured results, and the cost-transparency angle a monolith cannot offer.