This list covers image input: a photo, chart, or screenshot that you ask a model to explain. It does not compare image generation or claim a vision-quality rank. PocketLLM is pre-launch and focuses on text.
We checked the linked vendor sources on October 14, 2026. This post reports no comparative product test.
Before an image upload
- Remove private details from the image.
- Check the service policy for the selected account.
- Test a non-sensitive image with a known answer.
PocketLLM is pre-launch. Private, on-device AI, starting on iPhone, with more platforms planned. No account required, no in-app tracking, and chat processing stays on your device. Join the waitlist.
Claude, Gemini, Grok, Duck.ai, Llama 3.2 Vision and Qwen 2.5 VL
1. Claude
Claude supports JPEG, PNG, GIF, and WebP uploads. See the Claude file guide. Check the current account limits before a large task.
2. Gemini Apps
Google documents image and file uploads in Gemini Apps. See the Gemini upload guide. Check the account requirements and current limits.
3. Grok
Grok documents image uploads and account access through email, Google, Apple, or X. An X account is not the only method. See the Grok FAQ.
4. Duck.ai
Duck.ai supports image uploads with selected models. See its file upload guide. The privacy policy describes external model providers and optional data features.
5. Llama-3.2-11B-Vision-Instruct
Meta publishes image-input model weights and inference examples in its model card. Check runtime support, license terms, and memory before local use.
6. Qwen2.5-VL-7B-Instruct
Qwen publishes a vision-language model and inference examples in its model card. Check runtime support and memory before local use.
Claude, Gemini, Grok, Duck.ai, Llama Vision and Qwen VL: image checks
| Option | Check before use |
|---|---|
| 1. Claude | Check the file type and account limits. |
| 2. Gemini Apps | Check account access and file limits. |
| 3. Grok | Check the account method and file type. |
| 4. Duck.ai | Select a model with image support. |
| 5. Llama-3.2-11B-Vision-Instruct | Check the exact local configuration. |
| 6. Qwen2.5-VL-7B-Instruct | Check the exact local configuration. |
Conditions for local image inference
Local vision inference keeps images and prompts on the device during that inference. Local weights alone do not establish the data behavior of the app.
Local inference keeps prompts on the device during inference when the selected model runs there. For a local-only workflow, select no cloud model. Disable chat sync and cloud backup. Check the app telemetry policy. Disable optional network features. Review extensions, exports, and local storage access.
Check all labels, extracted numbers, and conclusions against the original image. Do not treat a plausible explanation as a verified result.
Frequently asked questions
Which image formats does Claude accept?
Claude supports JPEG, PNG, GIF, and WebP uploads. See the Claude file guide.
Does Grok require an X account?
Grok documents account access through email, Google, Apple, or X. See the Grok FAQ.
Does every Duck.ai model support image uploads?
Duck.ai supports image uploads with selected models. See its file upload guide.
PocketLLM is pre-launch, with a waitlist for the iPhone launch. More platforms are planned. See the PocketLLM model catalog for its separate model list.