Changelog

The evolution of the Sovereign AI ecosystem.

May 19, 2026

v0.3: The Reasoning Update

Latest

The catalog gets a full 2026 refresh and Stackend learns to host the new wave of open-weight reasoning models.

  • Reasoning models: Native support for DeepSeek R1 (7B / 32B distilled, MIT) and OpenAI's GPT-OSS (20B, Apache 2.0).
  • Frontier open weights: Llama 3.3 70B and Llama 4 Scout (10M-token context) added alongside Qwen 3 (8B / 32B) and Gemma 3 (4B / 27B multimodal).
  • Modern coders: Qwen 2.5 Coder (7B / 32B) replaces Code Llama as the default; Codestral 22B for multi-language work.
  • RAG-ready embeddings: First-class entries for nomic-embed-text and mxbai-embed-large.
  • Honest disclosures: Privacy page now spells out exactly what the optional onboarding API does (and doesn't) see.
December 7, 2025

v0.2: The AI Studio Update

A major leap forward in usability and management. This update introduced the "Stackend Studio" concept, giving you complete control over your local AI fleet.

  • Unified Dashboard: A real-time command center for model management, system telemetry, and health monitoring.
  • Model Lifecycle: Install, update, duplicate, and delete models directly from the UI.
  • Fine-Tune Ready: Added architecture for local model customization and fine-tuning.
  • Enhanced Reliability: Improved startup sequences and connection handling for a seamless experience.
November 2025

v0.1: Operation genesis

The foundation of Sovereign AI. We launched the core engine capable of turning standard bare-metal servers into AI-ready infrastructure.

  • Hardware Audit Engine: Advanced telemetry to detect GPU/CPU capabilities and match them with compatible models.
  • Smart Matchmaking: Automatic logic to determine "Optimal" vs "Insufficient" hardware compatibility.
  • Zero-Config Deployment: Automated orchestration of the entire AI stack with a single command.