Proprietary AI is often too expensive and centralized for many countries and companies. Yann LeCun argues that open-source AI is essential for achieving global AI sovereignty and cultural independence.

Apertus Mini consists of 16 small language models that showcase distillation and quantization techniques. Developed by the Swiss AI Initiative, the models are fully open with documented training data, code, weights, and alignment principles, and are designed to comply with EU AI Act requirements by respecting opt-outs, removing PII, and preventing memorization.

Local Qwen is a distinct AI tool, not inferior to Opus, with versions 27B and 35-A3B being compared to Opus level capabilities. Evidence from software businesses and open source projects supports this differentiation.

Z.ai has launched GLM-5.2, featuring a 1-million-token context window. The model will be released under an MIT license next week, promoting open access to frontier intelligence for global collaboration in artificial general intelligence (AGI) development.

TensorZero is an open-source LLMOps platform that provides a unified API for accessing various LLM providers with less than 1ms p99 latency. It includes features for observability, evaluation, optimization, and experimentation, allowing users to store inferences, benchmark workflows, optimize prompts, and manage experiments programmatically.

tastyeffectco/sandboxes provides a self-hosted development environment that allows users to create isolated cloud dev sandboxes with live preview URLs using a single command. The open-source engine supports rapid app development, enabling users to generate functional websites, such as todo apps, instantly on their own servers.

DeepSeek-V4 is now supported for both inference and reinforcement learning (RL) training from Day 0. SGLang and Miles provide the first open-source stack designed for DeepSeek-V4’s hybrid sparse-attention architecture and manifold-constrained hyper-connections, utilizing FP4 expert weights.

Thunderbolt is an open-source, cross-platform AI client that allows users to control and customize AI interactions. It is designed to be extensible, enabling developers to add new features and functionality.

Unsloth Studio is a no-code web UI that allows users to run and train AI models locally, supporting GGUF and safetensor models on Mac, Windows, and Linux. It enables training of over 500 models at twice the speed and with 70% less VRAM without sacrificing accuracy.

Key figures from Alibaba's Qwen AI team, known for their extensive contributions to open source generative models, have departed following the release of the Qwen3.5 small model series. The release received public acclaim from Elon Musk for its notable intelligence density.