Alibaba launches Qwen4-Max, and its own chart puts Claude Opus 5.5 ahead on 10 of 16 tests
Priced at $1.50 per million input tokens, 25% below its predecessor, Qwen4-Max ranks 38th of 81 on an independent index.
Priced at $1.50 per million input tokens, 25% below its predecessor, Qwen4-Max ranks 38th of 81 on an independent index.
Published software flaws doubled in a year. Google’s Project Zero is cutting its disclosure deadline, and fixes are struggling to keep pace.
South Korea’s top tax exam scorer says AI supplied his answers. The government is now auditing every licensing exam held in 2026.
The open-source vLLM server has published 58 security advisories since September, including several bugs that leaked data between users of one server.
OpenAI is moving every ChatGPT voice conversation to Live, the mode built on the GPT-Live model. Video and screen sharing keep the older engine.
Google says its agents finished porting the Zircon kernel to Rust in two weeks and made its video decoder faster than the original C++.
A Docomo survey finds Japanese grew less confident about spotting AI-made fakes between 2025 and 2026, in every age group.
A 30-hour outage by Microsoft’s count, Copilot included. Microsoft apologized, ruled out AI and promised a report within five business days.
Domyn is testing the setup for EUROPA, the open AI model picked by the European Commission, on the public JUPITER supercomputer.
US call centers now employ fewer people than at any time since 1994, and the decline began years before ChatGPT arrived.