⚡️ FlashAttention-4: up to 1.3× faster than cuDNN on NVIDIA Blackwell →

Introducing Together AI's new look →

🔎 ATLAS: runtime-learning accelerators delivering up to 4x faster LLM inference →

⚡ Together GPU Clusters: self-service NVIDIA GPUs, now generally available →

📦 Batch Inference API: Process billions of tokens at 50% lower cost for most models →

🪛 Fine-Tuning Platform Upgrades: Larger Models, Longer Contexts →

Cursor & Decagon at NVIDIA GTC

Name: Cursor & Decagon at NVIDIA GTC
Location: NVIDIA GTC

The best AI-native companies aren’t just picking the right models. They’re making hard calls on infrastructure, inference, and cost that compound over time.

Cursor and Decagon have been in the trenches building at production scale — and they’re sharing exactly what they’ve learned.

Hard-won lessons from two of the fastest-growing AI-native companies

How Cursor handles high concurrency inference for millions of developers

How Decagon scales enterprise AI agents without runway infrastructure costs

What it really takes to build production AI applications that hold up at scale

Enter your details to watch

conference
Under the Hood of Building and Scaling AI-Native Applications (Presented by Together AI
conference
Hard-Won Lessons From Production Inference at Scale (Presented by Together AI