Version note: Giskard is mid-transition to a major v3 rewrite (a modular, dependency-light architecture built for multi-turn agent testing), which is currently in beta/pre-release. As of this review, the flagship LLM Scan and RAGET features still run on the v2 engine and are being migrated to v3; v2 remains available but is no longer actively maintained. This is a genuine “which version do I adopt?” consideration for new users, and it’s reflected in the score below.

Giskard’s pricing follows the familiar open-core pattern, and its open end is genuinely generous. The open-source library — which includes the substance of the product: the LLM Scan, RAGET, the red-teaming scanner and the traditional-ML testing module — is completely free under the permissive Apache 2.0 licence. You install it with pip and run it locally against any model, with no usage fees and full data ownership, which makes it a real, self-contained testing solution for individual developers and teams, not a stripped-back teaser for a paid product. That’s the right place to start, and for many teams it’s all they’ll ever need. The enterprise Giskard Hub is where paid pricing comes in, and it’s positioned for organisations — particularly in regulated sectors like banking and insurance — that need the collaboration and governance layer: shared workspaces, role-based access control, versioned test-case management, custom failure categories, and, most importantly, continuous and scheduled red teaming that monitors production for new vulnerabilities automatically, along with the EU AI Act and OWASP compliance packs. Hub pricing is not published — it’s a contact-sales arrangement — so budgeting for it requires a conversation with Giskard, and the lack of a public number is a planning consideration if you have fixed budgets or long procurement cycles. Giskard also offers a professional LLM red-teaming service in which its own ML researchers conduct hands-on engagements to identify and mitigate vulnerabilities, useful for teams that want expert-led testing rather than only self-serve tooling. The practical guidance: start with the free open-source library to run scans and RAG evaluations locally and validate that Giskard fits your workflow (this costs nothing and reveals most of the value); adopt the Hub when you need team collaboration, continuous production red teaming, and compliance governance at organisational scale; and engage the red-teaming service for expert-led adversarial testing. As always, confirm current Hub pricing and terms directly with Giskard, and factor in the v2/v3 transition when planning which version to build on.