Krea has launched Krea 3, an image editing model the company describes as "built for production" — a system designed not to generate pictures from scratch, but to make precise, repeatable edits to existing images at a speed and price intended for commercial volume work.
According to Krea's launch page, Krea 3 completes a typical edit in under eight seconds and costs about $0.06 per 2K-resolution edit. The company positions those figures directly against Google's Nano Banana Pro and OpenAI's GPT Image 2.5, claiming Krea 3 is four times faster than Nano Banana Pro and, in its comparison table, listing a median request-to-result time of 26.8 seconds for Nano Banana Pro and 39.9 seconds for GPT Image 2.5 across a twelve-image test batch. For more context on this story, see our ongoing more AI stories.
The Benchmark Caveat Matters
Krea is upfront that its side-by-side comparison uses pre-generated outputs rather than a live benchmark: each model's listed time is the median of a single twelve-image batch, including queue time, and the company notes that the GPT Image 2.5 pricing figure of roughly $0.21 per edit is a "draft estimate pending validation." In other words, the headline numbers are Krea's own marketing measurements, not an independent evaluation — a distinction that matters in a market where speed and price claims are increasingly the primary competitive weapon.
Still, the price point itself is the story. At approximately $0.06 per 2K edit, Krea is pricing for pipelines that run thousands of edits per day — e-commerce catalogs, advertising variants, brand localization — where per-image cost and turnaround time dominate procurement decisions in a way that leaderboard rankings do not.
Editing That Touches Only What You Ask
The core technical claim of Krea 3 is restraint: the model "changes only what you ask" it to change. Krea's launch examples show regional edits where the changed pixels are measured against the original, alongside features aimed at production constraints rather than creative exploration.
Those features include hex-code color matching — name a specific color and Krea 3 applies it to the target product and only the target product — and native 2K output, which spares production teams an upscaling step. The company also demonstrated compositing that references multiple input images by number: seat a person from one photo at a specific location wearing a sweater from another, with a bag from a third placed on the counter. Pointing at people, products, places, and styles by number and having the model combine them into one coherent picture is aimed squarely at advertising and catalog workflows.
Character consistency across a sequence of edits is another emphasis: change the styling, setting, and mood of a scene while keeping a familiar face intact, which matters for campaigns that reuse the same model or mascot across dozens of assets.
From Generation to Editing
Krea 3's launch reflects where the AI image market's center of gravity has moved. The first wave of diffusion models competed on generation quality; the current wave competes on controlled editing, with Google's Nano Banana Pro and OpenAI's GPT Image line making iterative, instruction-driven editing a mainstream capability. Krea's wager is that professional editing workloads — where the source photo exists and only specific elements need to change — have different requirements than text-to-image generation, and that speed, price, and precision beat raw generative range for those customers.
The company's examples lean into commercial use cases: re-skinning a roadside billboard for a different brand, converting a textured illustration into a photograph from the same viewpoint, swapping a product in a street-style photo while preserving the rest of the frame. These are the kinds of tasks that currently occupy retouching teams, and Krea's pricing undercuts the per-image cost of the frontier alternatives by a wide margin on paper.
Whether Krea 3's quality holds up in independent testing — and whether its speed advantage survives real-world queue times at scale — will determine if the startup's production-first positioning converts into enterprise adoption. For now, the launch adds a third serious option to a category previously contested between two of the largest AI companies in the industry.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
