Qwen-MM-Plugins Releases

Performance optimizations for MoE models and prompt processing on Macs, plus fixes for model loading issues on various architectures.

Oct 3, 2026

Hardens security measures, reduces the build size, and refines the Cyber HUD with layout fixes.

Oct 3, 2026

Updates the default Text-to-Speech (TTS) provider in the pipecat-bot-py example to Deepgram Aura for significantly faster response times, and fixes a pause-mode barge-in bug where bot audio was dropped after resolution.

Oct 2, 2026

Introduces a new Pipecat transport (`siphon-ai-pipecat`) to integrate with Pipecat pipelines, ensuring real-time audio pacing and better handling of interruptions, along with a security update to Sightglass.

Oct 2, 2026

Addresses vision coordinate accuracy for specific AI models and updates the available model list by adding new entries and removing retired aliases.

Oct 2, 2026

This release selectively integrates upstream Laya runtime fixes into the native MLX port, focusing on conversation handling, validation, and language detection improvements.

Oct 2, 2026

Memory optimization and bug fixes for model loading, inference performance, and error handling on various platforms.

Oct 2, 2026

v0.40.0 introduces a new Skills tab for vaults, allowing users to add markdown instructions. The update also includes dependency bumps for React Router and Vite, along with a network fallback fix for NetGuard.

Oct 1, 2026

Bug fix release addressing Windows installer issues and improving error reporting for the inference engine.

Oct 1, 2026

Bug fixes for model assessment, tool call handling, image processing, and token counting to improve reliability across different models and platforms.

Oct 1, 2026

Refreshes the application interface and adds granular location data, including city-level search volume and scheduled rank checks.

Sep 30, 2026

This release focuses on fixing data structure handling for Excel files and document style ordering in Word files to ensure accurate schema and style application.

Sep 30, 2026

Adds a 'Windowizer' extension for creating parametric windows from wall faces, introduces visual selection indicators (dots), and improves performance for models with thousands of mirrored components.

Sep 30, 2026

Adds several new linting rules for React, Svelte, Astro, and CSS, along with significant performance improvements for formatting and analysis.

Sep 30, 2026

A massive update featuring 2,022 documented changes, introducing new providers like GPT-6 Astra/Sol/Luna and Xquik X, alongside significant improvements in routing resilience, video/audio bridge hardening, and security.

Sep 30, 2026

Integrates Blender rendering, adds an AI/MCP tab, implements a command search (F3), and introduces Extension API 2 with features like STL import and Paste in Place.

Sep 30, 2026

A major update focusing on Oracle backend compatibility, new visualization tools, and improved mental models. It includes the introduction of animated SVG figures, bank aliases, and per-scope consolidation strategies.

Sep 29, 2026

This release focuses on data integrity and 'grounding' numbers by identifying their origin (observed, derived, proposed, etc.). It includes a refactored final-answer grounding gate, improved backtest report grounding, and fixes for context window compaction. New markets (Argentina, Bahasa Indonesia) and broker connectors are added, along with Web UI improvements for configuring IM channels.

Sep 29, 2026

This release introduces relative partial references for self-contained partial trees, slug support for branch pages, and several security hardening fixes. It also includes improvements like automatic summaries that don't end inside containers and exact numeric comparisons in template functions.

Sep 28, 2026

Improves UI stability and enhances OAuth streaming capabilities across providers, along with refreshed model catalogs.

Sep 28, 2026