v1.0.1

What's Changed

Features

• High-performance LLM inference engine
• Support for Qwen3 models (8B/4B variants)
• Native embeddings and reranking
• CUDA support for NVIDIA GPUs
• Metal support for Apple Silicon
• Flash Attention v3 support

Installation

# Download for your platform
curl -L https://github.com/hanzoai/engine/releases/latest/download/hanzoai-$(uname -s | tr "[:upper:]" "[:lower:]")-$(uname -m).tar.gz | tar -xz
sudo mv hanzo-engine /usr/local/bin/hanzoai

Full Changelog: v1.0.0...v1.0.1

v0.8.0

What's Changed

Features

• High-performance LLM inference engine
• Support for Qwen3 models (8B/4B variants)
• Native embeddings and reranking
• CUDA support for NVIDIA GPUs
• Metal support for Apple Silicon
• Flash Attention v3 support

Installation

# Download for your platform
curl -L https://github.com/hanzoai/engine/releases/latest/download/hanzoai-$(uname -s | tr "[:upper:]" "[:lower:]")-$(uname -m).tar.gz | tar -xz
sudo mv hanzo-engine /usr/local/bin/hanzoai

v0.1.0 — Hanzo Engine

Initial release of Hanzo Engine — high-performance cloud inference engine.

Features

• 60+ model architectures (Llama, Qwen, Phi, Gemma, Mistral, etc.)
• CUDA 12+, Metal, and CPU backends
• Paged attention and continuous batching
• Speculative decoding and tensor parallelism
• OpenAI-compatible REST API
• Docker and Kubernetes native deployment

Links

• Documentation: https://engine.hanzo.ai
• Docker: ghcr.io/hanzoai/engine