v1.8.0 v1.9.0
cheaper-inference/OmniRouteFeb 22, 2026 → Mar 1, 20267 days between releases
What's New
Adds performance optimizations and multi-GPU support.
New Features
- Faster-Qwen3-TTS integration for CUDA acceleration.
- Multi-GPU assignment controls for subsystems.
- Trained model caching between generations.
- CUDA Graphs toggle in settings.
Improvements
- 5-10x faster inference for Qwen3 models via CUDA graphs.
- Automatic fallback to standard Qwen3TTSModel if CUDA fails.
- Automatic detection of multiple GPUs for assignment.
- Setup scripts updated to auto-install Faster-Qwen3-TTS.
Bug Fixes
- Fixed broken component reference in Conversation tab.
Upgrade Notes
Add new GPU assignment options to config.json. Update setup scripts to auto-install Faster-Qwen3-TTS. Ensure CUDA drivers are compatible if enabling CUDA graphs.
The Two Releases
Version 1.8.0 - Prompt Hub, Ollama Support & Split by Paragraph
Feb 22, 2026· 8 months ago
This update enhances the utility of the tools by adding a centralized Prompt Hub for quick access to saved prompts and introducing Ollama as a new LLM backend. It also adds a batch generation feature to split text by paragraph.
Full release notesVersion 1.9.0 - CUDA Graphs Acceleration & Multi-GPU Support
Mar 1, 2026· 8 months ago
This release prioritizes performance optimization by integrating Faster-Qwen3-TTS for CUDA graph acceleration and introduces multi-GPU support to allow subsystems to run on separate hardware.
Full release notes