Product / UX / Telemetry & observability

Galileo

AI observability and reliability platform combining offline evaluation, online production guardrails and tracing, driven by the vendor's own Luna-2 small judge models that detect hallucinations, tool errors and policy violations. Galileo's claim for Luna is that customized evaluations can monitor all production traffic at a fraction of the cost of a large judge model. It deploys as SaaS, into a customer VPC, or on-premises.

Galileo names no customers and publishes no usage figure, so the adoption band stands on an unquantified vendor claim and is directional. Verified 2026-08-13 via galileo.ai.

Openness

1 high confidence

1.0

license: proprietary
source: closed
deploy: SaaS/VPC/on-prem(managed)

A proprietary platform with no OSI-licensed core product, so it scores as closed.

https://galileo.ai/ recorded 2026-08-13
Homepage. The product is "our complete Agent Reliability platform", with proprietary Luna evaluation models that "monitor 100% of your traffic at 96% lower cost". No source repository, license or self-hostable distribution of the platform is offered anywhere on the page; the calls to action are Sign Up and Book a Demo.

Adoption

3 low confidence

3.0

Vendor states 'trusted by leading enterprises' but discloses no named customers or quantitative user/usage figures on the homepage. Level 3 reflects established commercial traction without a verified count; treat as directional.

https://galileo.ai/ recorded 2026-08-13
"Trusted by leading enterprises to measure, protect, and improve AI in production", and a second banner reading "Trusted by enterprises, loved by developers". Read for a figure and containing none - no customer count, no user count, no volume.

Capability

4 medium confidence

4.0

Strong coverage across the feature matrix: tracing that analyses agent behavior and failure modes, 20+ out-of-the-box evaluations for RAG, agents, safety and security plus custom ones, production guardrails, cost optimization through the purpose-built Luna-2 evaluation models at sub-200ms and around $0.02 per million tokens, and data capture across synthetic, development and production datasets. The loop from offline evaluation to runtime guardrail intervention and the Luna models are the differentiators. OpenTelemetry support is not confirmed on the homepage. That leaves it level with Langfuse at 4 rather than above the open source anchors.

https://galileo.ai/ recorded 2026-08-13
Homepage - customized "evals into Luna models that monitor 100% of your traffic at 96% lower cost", and the platform framed as measuring, protecting and improving AI in production, which is the offline-eval-to-runtime-guardrail loop the band rests on. OTel is still not named on the page.

Verified 2026-08-13