What's new

Updates worth knowing.

Release notes, product improvements, and important service announcements from Azure - kept in one place.

July 31, 2026

release July 31, 2026Latest

Analytics, security, and API platform expansion

A major platform update adds clearer usage insights, safer project handling, stronger API compatibility, and new service integrations.

Usage and cost analytics: The account dashboard now reports requests, tokens, latency, model usage, API-key usage, endpoints, and estimated model costs.
Admin and service health: Protected admin analytics now cover accounts, plans, projects, models, latency percentiles, failures, and provider health, alongside a public model-status page.
Plan expiry notifications: Plus and Pro expiry reminders are delivered in-app and can also be sent through email and configured webhooks.
Malware screening: Project writes, uploads, ZIP archives, previews, and project security scans now receive bounded malware and archive-safety checks.
More integrations: Encrypted Slack and Discord webhooks are supported, with secure GitLab and Bitbucket integration foundations.
Expanded developer platform: OpenAI Responses compatibility, token and cost estimation, JavaScript and Python examples, generated API documentation, Canvas snapshots, restore points, and resumable project jobs are now available.

July 30, 2026

release July 30, 2026

The model picker, Canvas workflows, and release notes are now easier to trust at a glance.

Model catalog cleanup: Permanently inaccessible models are removed from the web catalog, OpenAI endpoint, Anthropic endpoint, and attachment fallbacks.
One source of truth: Public model IDs stay consistent across chat and API clients, with private provider routing details kept out of responses.
Canvas continuity: Generated files, project context, previews, and follow-up edits stay connected to the active conversation.
What's new is now here: Release notes and service announcements live in one searchable page and appear once when a new update ships.

July 29, 2026

release July 29, 2026

Provider selection now favors the best available route without exposing internal workspace details.

Availability-aware selection: Unavailable routes are skipped before a request is sent, reducing dead-end responses.
Protocol compatibility: OpenAI-compatible and Anthropic-compatible clients share the same public model catalog and response identity.
Long-session context: Conversation context is carried forward with bounded compression when a provider cannot accept the full transcript.

July 27, 2026

release July 27, 2026

Multi-file work is organized as a real project instead of a loose collection of snippets.

Folders and files: Generated projects keep their folder paths, file names, and language metadata together.
Preview-ready output: HTML, CSS, and JavaScript files can be previewed together without flattening the project structure.
Safe revisions: Repeated edits target the selected file and preserve the rest of the workspace.

July 24, 2026

announcement July 24, 2026

Plan status, reset timers, API access, and promotional balances are visible from the account dashboard.

Clear plan access: The dashboard shows plan expiry, token reset countdowns, and the current API access tier.
Personal API keys: Pro accounts can generate and renew a personal key for OpenAI-compatible and Anthropic-compatible clients.
Promotion balances: Non-expiring promotion tokens are tracked separately and used after the regular allowance needs more room.

New releases will appear here and announce themselves once.