In partnership with
OpenAI just dropped two new models. One costs less than a cup of coffee per million tokens.
Today OpenAI released GPT-5.4 mini and GPT-5.4 nano — their smallest, fastest models yet. And the numbers are wild.
GPT-5.4 mini runs more than 2x faster than the previous mini, costs $0.75 per million input tokens, and still scores 88% on the toughest reasoning benchmarks. The full GPT-5.4 scores 93% on the same test and costs $2.50 per million inputs. You're giving up 5 points of performance for two-thirds of the price. For most tasks? That's the smarter trade.
The nano model is even more aggressive at $0.20 per million input tokens. That's basically noise. It's built for classification, data extraction, and the kind of boring-but-necessary background work that eats up quota when you use the big model for everything.
The bigger pattern here: OpenAI is explicitly designing these for multi-model workflows. The idea is that GPT-5.4 handles the thinking and planning, then hands off simpler subtasks to mini or nano agents running in parallel. We covered how Codex was already doing this with subagents last post. Now they're pricing the entire stack to make that architecture economically obvious.
The models that handle your most expensive work are getting cheaper. The ceiling is staying the same. The floor is dropping fast.
Every headline satisfies an opinion. Except ours.
Remember when the news was about what happened, not how to feel about it? 1440's Daily Digest is bringing that back. Every morning, they sift through 100+ sources to deliver a concise, unbiased briefing — no pundits, no paywalls, no politics. Just the facts, all in five minutes. For free.
Sam Altman said "quite the opposite" and the internet had opinions.
Someone on X called out the awkward math: OpenAI paid $6.5 billion to bring on legendary Apple designer Jony Ive, then reportedly pivoted away from consumer hardware without shipping a single product.
"Gotta know when to fold 'em but damn, 6.5 BILLION."
Sam Altman replied: "We are not shutting it down, quite the opposite. I think you will love what the team is building."
So what's actually going on? OpenAI is refocusing on enterprise customers and coding tools, scaling back moonshot projects like the browser and e-commerce features. The Jony Ive project isn't dead, just being reoriented around productivity rather than the original "ambient companion" vision. Anthropic has been eating their lunch in the enterprise market, and with an IPO reportedly on the horizon, investors want a tight story.
Nobody got a straight answer on what "quite the opposite" actually means. But you don't hire the guy who designed the iPhone and point him at a spreadsheet app. Something physical is probably coming.
$6.5 billion buys a lot of patience.
Google just made its most powerful Gemini feature free for everyone.
Personal Intelligence was previously locked behind Google's paid AI plans. As of today, it's available to all U.S. users with a personal Google account, across the Gemini app, AI Mode in Search, and Gemini in Chrome.
You connect Gemini to your Gmail, Google Photos, Maps, YouTube, and more, and it uses all of that context to answer you like it actually knows you. Planning a trip? It pulls your hotel confirmation from Gmail, checks your Maps history, and builds a full itinerary without you having to explain anything. No copy-pasting, no re-explaining your situation from scratch.
Google says you choose what to connect and it doesn't train on your personal data. But the bigger point is this: if you've been sleeping on Gemini, now is a good time to try it. The feature that made the paid plan worth it just became free.