Google's new image model is fast and cheap, and its chart omits ChatGPT
Nano Banana 2 Lite generates in seconds at 3.4 cents an image. An early tester noticed one thing missing from Google's comparison chart: any competitor that might win.
Google shipped Nano Banana 2 Lite, a distilled Gemini image model that renders in seconds at $0.034 an image. An early tester confirmed it behaves like a lighter Nano Banana 2 with good text rendering, but you can't force aspect ratios like the full model.
Two things the thread caught: the comparison chart quietly omits ChatGPT and DALL-E, and one longtime engineer noted Grok's image model beats Nano Banana on nearly every metric Google chose to highlight.
A distilled image model built for speed, not the top of the chart
Google DeepMind released Nano Banana 2 Lite, a stripped-down version of its Gemini image model pitched as its "fastest, most efficient" one yet and its "lowest cost yet," tuned for generating thousands of images cheaply and for real-time use inside apps and agents. The page claims roughly 2.7× faster generation than Gemini 3.1 Flash Image and leans entirely on side-by-side comparisons against its own full-fat predecessor, Nano Banana 2.
What the page doesn't do is put a number on the price or benchmark against rivals, and the thread filled both gaps.
Early testers supply the numbers Google left off
The most authoritative read came from someone with early access. A senior data scientist confirmed it is what it looks like, a distillation, with a specific limitation buyers care about:
"it does behave like a distilled Nano Banana 2 with respect to certain elements such as good text rendering… My main criticism is that you cannot programmatically force aspect ratios with NB2L but you can with NB2." — Max Woolf, senior data scientist at BuzzFeed
He pinned the price at $0.034 per image and called it higher than expected, since it generates in about half the time of a Nano Banana 1 image that costs $0.039 — and tied the whole launch to Google's move to offer free image generation in the Gemini app, which needs a cheap model underneath. An independent tester's numbers landed close: about 8 seconds and $0.0342 at 1K resolution, versus GPT Image 2 at $0.0414 and 35–45 seconds — faster and cheaper, though text comes out garbled and lighting skews dim unless you prompt around it.
Then came the omissions. One reader flagged that the comparison chart simply skips the obvious rival:
"They didnt include chatgpt in the comparison chart. That tells a lot" — an engineer who noticed the missing competitor
And even on the metrics Google did pick, a former computational-biology PhD turned early Twitch engineer pointed out the chart embarrasses its author:
"It's sort of amazing that Grok's image model beats Nano Banana on nearly every one of the metrics they chose to highlight." — timr, 31k karma
Not everyone was cynical: a reader who only dabbles in image gen was genuinely startled by 2-second generations that used to take a full minute on ChatGPT. But the sourest note was about consequences, not benchmarks — a real-estate gripe that cheap, fast interior generation means listings now bury the actual apartment behind a dozen AI-staged "Ikea-chic" fantasies before you can see the place. The faster and cheaper the model, the more of that the internet gets.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
An LLM wrote 100,000 lines putting Kubernetes in a browser; he read every one
Almost none of the code was written by a human. That's usually the setup for a disaster. Sam Rose's defense: he reviewed every line and wrote 2,000 tests asserting his browser cluster behaves exactly like a real one.