OpenAI DevDay 2026Codex in the CloudUltrafast Mode (8x)MicroVM SandboxesSafe CI/CD Deploys

OpenAI Codex in the Cloud & Ultrafast Sandboxes Studio

Orchestrate remote cloud programming environments where AI coding agents write, build, test, and safely deploy software. Harness Ultrafast Mode for up to 8x faster token generation loops inside isolated Firecracker microVM sandboxes.

Token Generation Velocity

8x Ultrafast

~380 tokens / second

Sandbox Isolation

MicroVM Firecracker

< 5ms boot cold-start

Loop Latency (Test-Build)

1.2s Fast Loop

Autonomous edit-eval cycle

Egress Security

eBPF Zero Trust

Restricted DNS & package lock

🏗️ Codex in the Cloud & MicroVM Sandbox Architecture

Ultrafast Mode Engine

⚙️ Sandbox & Engine Configurations

Generation Processing Tier

MicroVM Virtualization Engine

Safe Deployment Gate

💡 Developer Velocity & Security Impact

Harnessing Ultrafast Mode with Firecracker sandboxes compresses developer test-build feedback loops from minutes to 1.2 seconds, isolating arbitrary agent code execution while safely delivering verified pull requests.

Loading Codex in the Cloud configuration...