This report presents the Qwen3-ASR family, featuring two advanced speech recognition models that support 52 languages. The 1.7B model offers top performance among open-source options, while the 0.6B model balances accuracy and efficiency, achieving rapid transcription and efficient forced alignment for text-speech pairs. Both models are released under the Apache 2.0 license for community use.

LLMc is a novel compression engine that utilizes large language models (LLMs) to achieve superior data compression by leveraging rank-based encoding. It surpasses traditional methods such as ZIP and LZMA, demonstrating enhanced efficiency in processing and decompression. The project is open-source and aims to encourage contributions from the research community.

Notte is a web agent framework designed to enhance the efficiency and reliability of building and deploying AI agents that interact with websites. It allows for a hybrid approach that combines traditional scripting with AI, resulting in significant cost savings and improved performance, as well as offering features like stealth browser sessions, credential management, and digital personas. Developers can utilize a single API to create custom web automations and agents while leveraging tools for structured output and secure data handling.