by Raphael Dudler | Jul 6, 2026 | gafam watch
On June 30, Chinese food-delivery giant Meituan open-sourced LongCat-2.0 — a 1.6-trillion-parameter model released under a permissive MIT license, which the company says is the first model of its scale trained and run entirely on domestic Chinese chips, with no Nvidia GPUs. The benchmark claims are self-reported and the full weights aren’t yet public. But the hardware milestone is the story — and it lands as Washington gates access to American models. gafam.ai reads what it means for Europe’s impossible AI choice.
by Raphael Dudler | Jun 24, 2026 | gafam watch
OpenAI expanded its Daybreak cybersecurity programme on June 22, releasing the full version of GPT-5.5-Cyber — its most capable vulnerability-finding model, scoring 85.6% on CyberGym — alongside “Patch the Planet,” an open-source security initiative with Trail of Bits and HackerOne. It is a direct competitive answer to Anthropic’s Project Glasswing. But the launch carries a warning that should concern every European organisation: AI now finds security flaws faster than humans can fix them. gafam.ai’s European analysis.
by Raphael Dudler | May 21, 2026 | meta ai
Meta has launched Muse Spark — its first flagship large language model built entirely in secret, under Chief AI Officer Alexandr Wang’s newly formed Superintelligence Labs. It is a closed, proprietary model — a direct departure from the open-source Llama strategy that Meta has championed since 2023. Muse Spark delivers competitive performance on multimodal perception, reasoning and agentic tasks at a fraction of the compute cost of Llama 4. The open-source era at Meta may be ending.
