Your image remains in browser memory until you explicitly run the test.
PHASE 07 / VISION PLAYGROUNDOne image.
One image.
Multiple model eyes.
Test how supported multimodal models interpret the same image and question.
Upload locally, control the analysis instructions, run a protected live comparison, then inspect answers, latency, tokens, and cost.
Every selected model receives the same image, prompt, and limits.
Inspect output, speed, tokens, and provider-reported cost side by side.
Review visual claims against the source image before relying on them.
What should the models inspect?
Use a non-sensitive JPEG, PNG, or WebP image under 4 MB. AXON//RADAR does not store the image.
Add one image.
The preview and API key stay in memory only. The image is sent to selected providers only when you press Run.
Ask the same visual question.
Compare supported models.
Only models explicitly marked for image input are available here.
MODEL CONNECTIONUSE YOUR OPENROUTER KEY
Enter your own OpenRouter key. It is held only in this tab and is never saved.
READY TO ANALYZE
Gemini 3.7 Flash vs Qwen3.8 27B
Uploading creates a local browser preview; it does not upload the file by itself. Running sends the image and prompt through AXON//RADAR’s protected endpoint to each selected model provider. Files are type-checked, limited to 4 MB, never stored by AXON//RADAR, and excluded from caching. Do not submit confidential images, identity documents, medical records, private addresses, or material you lack permission to process.