Galileo Pricing Explained (2026) - What You Actually Pay

Galileo has the most generous free tier of any enterprise eval platform, and the cleanest paid tier to understand - right up until the point where it stops being self-serve. Let me decode the meter, walk a worked bill, then point at two cheaper picks if the cliff past Pro is a problem.

First, the disambiguation, because it poisons every pricing page you will find. There are two unrelated companies named Galileo. The one here is the LLM eval platform at galileo.ai. A different text-to-UI design tool at usegalileo.ai was acquired by Google. If you read a “$39 a month” plan anywhere, that is the wrong company. Everything below is the eval platform, verified against galileo.ai.

The pricing model, decoded

Galileo has two self-serve tiers and then a sales call.

Tier	Price	Traces/mo	Deployment
Free	$0	5,000	Cloud
Pro	$100/mo billed yearly	50,000	Cloud
Enterprise	Contact sales	Unlimited	Cloud, VPC or on-prem

The quota unit is traces per month. The free tier is genuinely good - 5,000 traces, unlimited users, unlimited custom evals. That is enough to run a real evaluation, and unlimited seats at $0 is unusual in this space.

Pro is $100/mo, and the important word is “billed yearly.” The page markets a 33% annual saving, so the real commitment is $1,200 up front for the year, not $100 you can cancel next month. Pro gets you 50,000 traces, standard RBAC, advanced analytics and a dedicated Slack channel.

The trap is the cliff above Pro. The page says pricing “scales based on number of traces” past 50k, but it does not tell you the rate. Everything that makes Galileo an enterprise platform - unlimited traces, VPC or on-prem, SSO, real-time guardrails at scale, a dedicated CSM - is Enterprise contact-sales. So you can size your bill precisely up to 50,000 traces a month, and not one trace further without a rep on a call.

Estimate your bill

Two clean scenarios, then the fog.

Prototype or small eval loop. Under 5,000 traces a month costs you $0. If you are validating whether Galileo’s Luna eval models catch your hallucinations, you never touch the paid tier. This is the case Galileo makes easy on purpose.

A team running real evals. Up to 50,000 traces a month is $100/mo billed yearly, so $1,200 for the year. RBAC and advanced analytics come with it. Predictable, and cheap for an enterprise-grade eval platform.

Anything bigger. Say you are tracing 200,000 requests a month and want guardrails in production. There is no public number for this. You are in a sales conversation, and the honest answer is you cannot forecast the bill from the pricing page. Budget time for a call, not just money.

The Luna angle, and why the free tier exists

Galileo’s bet is its proprietary Luna and Luna-2 eval models - small models fine-tuned for tasks like hallucination and prompt-injection detection. The pitch is that they are cheap and fast enough to run on every request, which turns offline evals into real-time guardrails. Galileo claims up to 11x faster and 97% cheaper than a GPT-3.5-based judge, but those are vendor benchmarks, not independent measurements - treat them as a claim until you run them on your own traffic.

That is also why the free tier is generous. Galileo is the best-funded platform in the space at roughly $68M raised, including a $45M Series B in October 2024, so it can afford to give away 5,000 traces to get you hooked on the eval quality before the enterprise motion starts.

Cheaper alternatives if the cliff hurts

If you outgrow the 50k-trace Pro tier and do not want a sales cycle, two open-source platforms undercut Galileo and let you self-host for free.

Langfuse is the open-source observability default. It is MIT-licensed and self-hosts free, with only three features enterprise-gated. Managed, it is $29/mo for its Core tier, and at 1M events a month it runs about $101/mo - a fraction of what a sales-led enterprise contract costs. The catch is operational, not commercial - the v3 self-host needs Postgres plus ClickHouse, Redis and S3-compatible storage, four services to run. But you can forecast the bill to the dollar, which Galileo past Pro will not let you do. The full breakdown is in Langfuse pricing.

Opik is Comet’s platform, and its cloud is the cheapest here - $19/mo for 100k spans, with $5 per additional 100k. The OSS build is Apache-2.0 with the full feature set self-hosted, no gates and no caps. Note the meter differs - Opik bills spans, Galileo bills traces, and one trace can contain many spans - so compare on your own workload, not the headline. But for a team that wants transparent, card-payable pricing with no enterprise call, Opik is hard to beat.

So which one?

• You are an enterprise burned by production hallucinations and you want research-grade evals plus guardrails - Galileo, and go in knowing everything serious lives behind a sales call. Confirm the domain first.
• You want the same job with a bill you can forecast - Langfuse if you can run the self-host stack, Opik at $19/mo if you would rather stay managed.
• You are still under 5,000 traces a month - stay on Galileo’s free tier. It is the most generous entry point in the category, and you lose nothing by testing the Luna models there first.

For the full field of options past the 50k cliff, see our Galileo alternatives and the best LLM observability tools rankings. Every price here was read from galileo.ai on 26 July 2026. This category ships breaking changes monthly, so we re-verify every 30 days.