Gemini Omni 1.2 Flash Makes 15-Second Clips, 30% Cheaper
Longer clips, chat-style edits and product references that hold their shape, plus three free videos a month in Vids.
AI tools, models and search for website owners.
Longer clips, chat-style edits and product references that hold their shape, plus three free videos a month in Vids.
OpenAI says it found 81 new Firefox flaws and built 948 Linux kernel exploits at about $1,300 each. Those are its own numbers, so far.
A five-person lab says its 65B model beats open rivals with 500B or more parameters on TerminalBench. The scores are its own, so test before you switch.
The API price stays where the preview was, and the benchmarks are Tencent's own, so test it on your work before you switch.
Apple staff can use Claude, ChatGPT and Gemini, but some teams are now told to pick cheaper or on-device models. Other big companies are rationing too.
Uber says every AI company is talking to it, but almost nobody is booking through an agent yet. Here's what that means if you're building for agents.
It's research, not a product: no lab plans to run it yet, and it only covers short answers without randomness.
Conductor's team joins Meta Superintelligence Labs, and its cloud service shuts on November 20, with user cloud data deleted after that.
You can now change a GIF's speed and frames inside ChatGPT, and OpenAI says images come out 50% faster.
Ahrefs data shows ChatGPT still sends websites less than half a percent of tracked traffic.
If your app has a companion that flirts, even as one feature among many, it needs a new App Store Connect declaration.
Three more newspapers filed a similar suit the same day, and the Times says publishers have no way to block Brave's crawler.
If you publish text drafted with Claude, your older models are about to mark it too.
If your team sends ChatGPT share links to clients, the link itself was the attack, and the fix needs nothing from you.
The nonprofit behind Wikipedia takes neither side, and draws the same line publishers care about: the answer box, not the training run.
Siri can hand some requests to Qwen, and the new Siri AI from WWDC 2026 isn't part of the first release.
It's a research prototype, and its agent sessions run with full access as your own OS user.
Instant chat stays off the meter, but OpenAI hasn't said how big each plan's shared allowance will be.
For SaaS and service businesses, it's a new way for customers to use your service without visiting your site.
The cheaper model is built for screen work like forms, drawings and video edits, and it's aimed squarely at cost.
Moonshot's new model scores close to Claude Opus 5.5 on a coding agent test at about half the token price.
Your old integration keeps working until you upgrade the SDK, which is exactly when it can break.
It's built for professionals and small business owners, and every send, payment or deletion still needs your approval.