Account and Settings

The Account page is where you manage your Testzilla account: your credit balance and plan, billing, your API key, and the Evaluation model that judges your tests. Open it from the Account entry at the bottom of the left nav in the web app.

What’s on the page

• Email and balance — your sign-in email and current credit balance.
• Plan and credits — your current subscription plan, credits remaining and renewal date, and options to upgrade, downgrade, or buy more credits.
• Billing — Transaction History and Manage Billing.
• API Key — your account API key for API and MCP access (see Authentication).
• Evaluation model — the model that decides each test’s pass/fail/uncertain verdict, covered below.
• Delete Account — permanently removes your account (your balance must be zero first).

Evaluation model

Every test you run is judged by an LLM that reads the transcript and returns a verdict — pass, fail, or uncertain — against your test’s criteria. The Evaluation model setting controls which model does that judging for your account. It is a per-account choice and applies to all of your tests.

There are two options:

Option	What it is
Default (GPT-4o)	The platform default evaluator. Recommended, and what every account uses unless you change it.
Claude Sonnet	Anthropic’s Claude Sonnet, run on AWS Bedrock in the EU region (Zero-Data-Retention-capable). An opt-in override for data-residency-sensitive customers.

Pick Default (GPT-4o) unless you have a specific reason to switch. Changing the evaluation model does not change your tests, criteria, or how results are displayed — only which model produces the verdict.

Why choose Claude Sonnet (EU / Zero-Data-Retention)

The Claude Sonnet option runs on AWS Bedrock in the EU region (eu-west-1) rather than a US-based API. Requests are routed within the EU (using EU cross-region inference profiles), and the deployment is configured to be Zero-Data-Retention-consistent: model-invocation logging is disabled in-region, so prompts and completions are not retained.

This is aimed at EU and regulated customers who need their evaluation traffic to stay in-region for data-residency and trust reasons. If your test transcripts may contain personal or sensitive data and you are subject to GDPR or similar requirements, the EU / Zero-Data-Retention path keeps that evaluation step inside the EU.

Capability, not a certification

We describe this option as Zero-Data-Retention-capable and EU-region, not as a formal certification. ZDR and EU routing are configured at the AWS account and region level. If you have specific compliance requirements, talk to us before relying on it for production data.

May need enabling for your account

The Claude Sonnet option only appears as selectable when the EU / Bedrock path is provisioned for your environment. If it shows as “Contact admin to enable,” your default GPT-4o evaluation keeps working unchanged — reach out and we’ll switch the EU evaluator on for your account.

How to change it

1. Open the Account page.
2. Under Evaluation model, click the option you want — Default (GPT-4o) or Claude Sonnet.
3. The choice saves immediately and applies to your subsequent test runs.

Switching back to Default (GPT-4o) at any time clears the override and returns you to the platform default.