Local iPhone AI (WebGPU)
Canvas Preview
Qwen2.5 1.5B
Model
SmolLM2 360M (Very Lightweight)
Qwen2.5 1.5B (Smarter Local)
Gemini 3.1 Flash API (External)
Execution Mode
Lite (256 tokens)
Lite + Validation (1 Check)
Fast (Search Grounded, 256 tokens)
Fast + Validation (2 Checks)
Deep Thinking (2,500 max tokens)
Deep Thinking + Validation (3 Checks)
Ultra (10,000 tokens, 2 Checks)
Max (100,000 tokens, 5 Checks)
Research (5-50 pgs, 50k tokens)
Deep Research (Book + Images, 100k tokens)
Custom API Config
Custom OpenAI-Compatible API Configuration
š Select your desired model and execution mode above. Code generated in chat will automatically render in the live Canvas Preview!