- GLM-5.3-Flash a 320B-parameter multimodal MoE model with about 18B active.
- Ships with 1M-token context and MIT-licensed weights.
- Reports say it is being served on domestic Chinese accelerators.
- Strong early usage and benchmarks put cost pressure on Western closed APIs.
- New 360p draft mode is faster and cheaper; finished output can upscale to 1080p or 4K.
- Available now via Gemini API and Google AI Studio for production-style video workflows.
- Omni 1.1 Flash adds scene extension using up to 10 seconds of prior context, supporting clips up to about 40 seconds total.
- Developers get first/last-frame control for smoother transitions and camera moves.
3/3 — The control battle is shifting to the phone. Apple and Google are clashing with the EU over AI assistants, where default placement and OS access could decide winners. The next AI monopoly fight may be less about model IQ than distribution.
Tech’s week in one line: AI moved from “chatbot race” to infrastructure, security, and control-of-the-internet battles. The stakes are now measured in gigawatts, vulnerabilities, and platform power. #TechNews#AI#Innovation
2/3 — AI infrastructure just got a $9.8B reality check. Hut 8 signed a 15-year AI-data-center lease covering 352 MW in Texas, taking its contracted AI capacity to 949 MW. The AI boom’s scarce asset isn’t only GPUs—it’s power, land, and grid access.