Features
Native PyTorch eager execution on TPUs
Fused Eager mode (50-100%+ speed gains)
Distributed training (DDP, FSDP)
Mixed precision training with FP8 on Ironwood TPUs
Integration with PyTorch Lightning, Hugging Face Transformers
Zero static graph compilation required
Scale to 100K+ chip clusters
Open-source backend (torch-xla) on GitHub
XLA compiler integration for optimized performance
Training and inference with vLLM on TPU
Model serving with vLLM unified backend (JAX & PyTorch)
Compatibility with existing PyTorch codebases
Supports Gemma 4 inference on vLLM TPU
Integration with MaxText for LLM training
Integration with Metrax metrics library (JAX)
Visual screen editor with infinite canvas
AI agents: Claude, GPT, Gemini, Qwen 3.7 Max, Claude Opus 4.8, Gemini 3.5 Flash
BYO API keys for OpenAI, Anthropic, OpenRouter on any plan
Live preview on device via QR code
Built-in code editor for pixel-level control
One-click publishing to web, iOS, Android
Full version control with rollback
Shared task board for organizing work
Team collaboration with roles and permissions
REST API integration with one-click setup (Webflow, Shopify, Xano, Supabase)
Supabase backend (database, auth, storage) in open beta
Theme and typography editing with Google Fonts picker
MCP integrations: Cloudflare, HighLevel
Exportable source code with full GitHub history
Expert services: scoping, design, backend, troubleshooting
