Luma AI Challenges Google with Uni-1: A Serious Rival to Nano Banana

Uni-1 from Luma AI: A Serious Competitor to Google's Dominance

Update from March 23, 2026

The Uni-1 model from Luma AI is now publicly accessible. According to human preference test results, known as Elo evaluation, Uni-1 stands out as the leader in several categories, including overall generation, style and editing, as well as reference-based generation. However, in the category of image generation from text, it ranks just behind Google's Nano Banana.

In personal tests, Uni-1 impressively responded to a reference prompt, competing with the Nano Banana Pro and even surpassing the new Midjourney v8, which showed weaknesses with the same prompt. It is important to note that the image produced was generated by a Luma agent, meaning the results may slightly differ from those obtained via the upcoming official API. Users can currently test Uni-1 for free on the Luma Labs platform.

Prompt: A hyper-realistic DSLR photo. A monkey holding a pink banana is sitting on a tiger in the foreground. In the background, a HORSE is riding an ASTRONAUT. The astronaut is underneath like a living "space saddle," and the HORSE is clearly on top, in control, as the rider. Make this 100% unambiguous: the HORSE is the rider and the ASTRONAUT is mounted, not the other way around. High resolution, sharp focus, realistic lighting. (Best image out of three attempts, but all three were good...)

Overall, Uni-1 is closing in on Google's flagship model while offering a lower cost for equivalent resolution. At a resolution of 2K, the average cost via the upcoming API is about $0.09 per image, depending on the number of reference images provided.

• Text to image generation (2048px)
• Image editing / i2i (2048px)
• Multi-references, 1 image (2048px)
• Multi-references, 2 images (2048px)
• Multi-references, 8 images (2048px)

In comparison, Google's Nano Banana 2 offers lower resolutions at more competitive rates: an image of 0.5K costs about $0.045, and an image of 1K around $0.067.

Original Article from March 8, 2026

Luma AI recently unveiled Uni-1, its first model integrating both image understanding and image generation within a unique architecture.

Like Google's Nano Banana Pro and GPT Image 1.5, Uni-1 is based on an autoregressive transformer. This AI model generates content sequentially, token by token, unlike traditional diffusion models that create images from noise. Text and images are processed in a common pipeline.

Luma claims that this model is capable of reasoning through prompts, both before and during generation, by breaking down complex instructions and planning scenes. This method allows for more precise prompt tracking, and Uni-1 is no exception. For example, it can combine multiple photos to create a new composition.

Ordinary photos of pets were merged into the scene described above. Prompt: "Combine the black and white curly-haired dog with a pink bandana, the Boston Terrier in a plaid harness, and the black and white cat into a single scene where they are dressed in academic attire, standing in front of a whiteboard filled with scientific diagrams and text, with the Luma AI logo placed in the upper left corner."

Beyond basic generation, Uni-1 can refine subjects through multiple exchanges while keeping the context intact. It is capable of converting images into over 76 artistic styles, accepting sketches and visual instructions as input, and transferring identities, poses, and compositions into new images from reference photos. In a demonstration, the model generated an entire sequence from a single reference image, gradually aging a pianist from childhood to old age.

According to Luma, Uni-1 scores the highest on the RISEBench test for logic-based image processing, narrowly beating both the Nano Banana 2 and GPT Image 1.5. The image generation capability also enhances the model's visual understanding. In object recognition, for example, it nearly matches Google's Gemini 3 Pro. The model supports multiple languages.

Uni-1 ranks at the top of the overall RISEBench leaderboard, just ahead of the Nano Banana 2 and GPT Image 1.5, the current image model powering ChatGPT.

Uni-1 will soon be available via Luma Agents, a newly launched creative assistant, and the Luma API. No pricing has been announced yet.