shimmy Releases

Introduces native SafeTensors support with zero Python dependencies and improved model detection.

Sep 10, 2025

Focuses on code quality and testing with the implementation of a property-based testing framework.

Sep 9, 2025

Critical bug fixes addressing model discovery, OpenAI API compatibility, and performance issues.

Sep 9, 2025

The first stable production release of Shimmy, featuring comprehensive cross-platform support, a mature OpenAI API compatibility layer, and a production-ready inference engine. It also introduces enhanced Ollama model discovery and automated build infrastructure.

Sep 8, 2025

Adds native Ollama integration and enhances community support infrastructure.

Sep 6, 2025

Initial release of a lightweight 5MB Ollama alternative with full OpenAI API compatibility.

Sep 4, 2025

v0.6 introduces improved language understanding, particularly for Hindi, and adds noise robustness features. It supports new model variants (Gemma3, Qwen3) and integrates VoiceBench evaluation support.

Aug 18, 2025

v1.23.0 focuses on UI improvements, including a restart option and Windows Arm builds. It also updates Qt and improves input capture portal functionality.

Jul 23, 2025

Ultravox v0.5 delivers significant performance improvements in transcription accuracy and translation capabilities, expanding language support to 42 languages and introducing new evaluation metrics.

Feb 11, 2025

Minimal release notes: "### Patch Changes - #1326 172fb0a Thanks @Schniz! - fix --use-on-cd failing with newly released default of --resolve-"

Nov 17, 2024

This release marks the stabilization of the `--resolve-engines` flag by enabling it by default. It also updates macOS binaries to be universal executables and introduces install/uninstall aliases.

Nov 14, 2024

This patch release upgrades the Whisper encoder to Whisper-large-v3-turbo and adds six new languages, bringing the total supported languages to 15.

Nov 12, 2024

A patch release focused on bug fixes including shell type handling, download fallback mechanisms, and performance optimizations for the `fnm env` command.

Oct 6, 2024

Ultravox v0.4 introduces a Whisper-medium encoder and trains on a larger multilingual dataset compared to v0.3, enabling support for 9 languages.

Aug 27, 2024

Ultravox v0.3 introduces a Knowledge Distillation approach for loss calculation and synthetic data augmentation to improve speech-to-text alignment.

Aug 23, 2024

This is a patch release that resolves an issue with shell inference on Windows. The update fixes a bug where PowerShell was failing to correctly infer the shell when using Scoop's shims.

Jun 13, 2024

May 16, 2023