Draw Things
Draw Things 1.20260716.0 is an AI image-generation app for iPhone, iPad and Mac. It can run models locally on Apple devices and also offers optional cloud or server offload paths.
What Draw Things provides
Draw Things brings local AI image generation to iPhone, iPad and Mac. After a model is available on the device, local inference can reduce network requests, while model downloads, updates and optional cloud or server offload have separate storage and privacy boundaries.
Models and device limits
Model size, quantization, memory and chip architecture affect startup time, image size and batch generation. Begin with a small model and low resolution, keep model and LoRA files organized, and record the model, seed and settings with exported images.
Maintenance note
This page reviews Draw Things 1.20260716.0 for iPhone, iPad and Mac, including model setup, local generation, storage, permissions and cloud-processing boundaries. Content review date: 2026-08-23.
Save to your cloud drive
Open the cloud drive to get the file directly, or save it for convenient access on another device.
Quark Cloud Drive
RecommendedSave Draw Things to this cloud drive
Baidu Netdisk
Save Draw Things to this cloud drive
Draw Things 1.20260716.0 local image generation guide
Install the app, download a model suitable for the Apple device, and create one low-resolution test image before changing advanced settings.
Before you start
- Confirm the system version and reserve space for models, cache and generated images.
- Use non-sensitive prompts and authorized source images during the first test.
- Decide whether the current task uses local inference, cloud inference or server offload.
Installation steps
- 01
Install and check storage
Install through the normal Apple device flow, then review available storage and photo or file permissions and grant only what is needed.
- 02
Choose a model for the device
Start with a smaller model that the device supports, avoiding several large downloads until memory and storage behavior are known.
- 03
Confirm the computation path
Check whether generation is local or remote, then disable accounts and cloud features that are not required for the test.
Quick start
- 01
Send a simple prompt
Use a short prompt, the default negative prompt and a low resolution with one output so model loading can be checked.
- 02
Compare one parameter at a time
Keep the model and seed fixed while changing one setting such as steps, guidance or size, making the result easier to understand.
- 03
Save the generation record
Export the image together with model, seed, prompt and key parameters, then review metadata and license boundaries before sharing.
Usage tips
- Models and LoRA files consume persistent storage; remove them only after confirming that no project depends on them.
- High resolution, batches and training increase heat and memory pressure, especially on mobile devices.
- Local generation does not imply every feature is offline; model retrieval and optional remote paths still use the network.
Troubleshooting and uninstall
Why does a model make the app close?
Free memory and storage, try a smaller or more heavily quantized model, and lower resolution. Older devices may not run current large models.
Why does a model download fail?
Check network, free space and system time, remove the incomplete cache and retry through a known source rather than importing an unverified replacement.
- Export images and settingsSave required images, prompts and model records, then confirm that important work has an independent copy.
- Remove models and the app separatelyDelete large models and caches only after checking project dependencies, then uninstall the app through the normal Apple device flow.
Frequently asked questions
Can Draw Things generate images without an internet connection?
Yes, when a compatible model is already downloaded and local inference is selected. Model retrieval, updates and remote features still need a network.
Will an older iPhone or Intel Mac run every model smoothly?
Installation support does not guarantee a good experience. Memory, system version, chip architecture and quantization determine the usable model and image size.
Where should model files be stored?
Keep models and LoRA files in the app-managed or documented location, reserve enough storage and record the path before migration or cleanup.