Apple is developing Manzano, a new multimodal model designed to perform both image understanding and image generation within a single system. This dual capability represents a significant technical challenge that has historically limited most open-source models, which often excel in one area but fall short in the other. Apple says Manzano addresses this by integrating both functions seamlessly, matching the performance of commercial systems like OpenAI’s GPT-4o and Google’s Nano Banana. Manzano, named after the Spanish word for "apple tree," has not been released publicly and does not yet have a demo. However, Apple researchers have published a paper showcasing low-resolution image samples generated from complex prompts. These results are compared against outputs from open-source models such as Deepseek Janus Pro and commercial systems like GPT-4o and Gemini 2.5 Flash Image Generation. In evaluations using three demanding prompts, Manzano delivered results competitive with GPT-4o and Google’s Nano Banana, particularly in tasks involving dense text, such as reading documents or interpreting diagrams—areas where many models struggle. The core issue Apple identifies lies in how models process images. Understanding images works best with continuous data streams, while image generation requires discrete tokens. Most models use separate pathways for each task, leading to internal conflicts within the language model. To overcome this, Manzano employs a hybrid image tokenizer. This system uses a shared image encoder to generate two types of tokens: continuous tokens for comprehension and discrete tokens for generation. Because both streams originate from the same encoder, the mismatch between tasks is minimized. The model’s architecture consists of three key components: the hybrid tokenizer, a unified language model, and a dedicated image decoder. Apple developed three versions of the decoder with 0.9 billion, 1.75 billion, and 3.52 billion parameters, supporting image resolutions from 256 to 2048 pixels. Training occurred in three stages using 2.3 billion image-to-text pairs from public and internal sources, along with one billion internal text-to-image pairs. The total training data amounted to 1.6 trillion tokens, including synthetic data from systems like DALL-E 3 and ShareGPT-4o. On benchmark tests, Manzano outperformed other models in several key areas. The 30-billion-parameter version achieved top results on ScienceQA, MMMU, and MathVista—tasks that require strong text and diagram comprehension. Performance improved steadily as model size increased from 300 million to 30 billion parameters, with the 3-billion-parameter version scoring over 10 points higher than the smallest version on multiple tasks. Manzano 3B and 30B ranked at the top of nine multimodal benchmarks, demonstrating strong performance across both understanding and generation. When compared to specialized models, the performance gap was minimal—less than one point for the 3-billion-parameter version. On image generation benchmarks, Manzano ranked near the top, successfully handling complex instructions, style transfer, image editing, inpainting, outpainting, and depth estimation. Apple views Manzano as a promising step toward more flexible, modular multimodal AI. Its design allows individual components to be updated independently and integrates techniques from various AI research areas. However, despite these advances, Apple’s foundation models still lag behind industry leaders. To bridge the gap, the company plans to incorporate OpenAI’s GPT-5 into Apple Intelligence starting with iOS 26. While Manzano shows strong technical progress, its real-world impact will depend on future updates and integration into Apple’s ecosystem.