Week of February 3: GPT-5.2 Codex, TEE Attestation UI, and Subscription Reliability

This week focused on giving you access to stronger models and making sure the subscription system works flawlessly.

GPT-5.2 Codex is live

GPT-5.2 Codex is now deployed and available in VibeBrowser. This is OpenAI's latest code-focused model, and it excels at complex browser automation tasks — especially those involving forms, multi-step workflows, and data extraction.

We consolidated our Azure OpenAI resources to a single region for lower latency, so you should notice faster response times across all Azure-hosted models.

TEE attestation in the model selector

When you pick a TEE-enabled model, VibeBrowser now shows attestation verification directly in the model selector. A green badge confirms the model is running inside a verified secure enclave, so you know your data is protected before you start.

Credential store improvements

The credential manager got better testing and documentation. If you store login credentials for sites the agent visits frequently, it now handles them more reliably — especially for sites with complex authentication flows.

API health monitoring

A new /api/health endpoint lets you check the status of VibeBrowser's backend services at a glance. This is useful for teams running their own deployments and for our own monitoring infrastructure.

Subscription system hardening

Several fixes made the subscription flow more robust:

• Tier unlock on login — Max tier models now unlock immediately after authentication, not on the next page load
• Trialing subscription support — the Stripe integration now correctly handles trial periods
• Graceful fallback — if the usage API is temporarily unavailable, the extension continues working instead of blocking

Under the hood

• Unified test authentication API across all E2E test suites
• Added GitHub Copilot subscription testing
• Fixed CI artifact paths and mock server limits
• Improved LiteLLM rollout timeout for large deployments