Plans & pricing

Every plan starts free — no credit card. You can run real, statistically defensible experiments on the free tier; paid plans add LLM-as-judge quality grading, team seats, a bigger event allowance, and longer retention. The live pricing and enterprise interest form live at llmjury.com/pricing.

Plans at a glance

PlanPriceEvents / moSeatsLLM-as-judgeRaw-event retentionMax permutationsMax sample / arm
Free$050,00017 days1,00010,000
Pro$149 / mo1,000,0001050k evaluations / mo30 days2,00050,000
Business$799 / mo10,000,00050500k evaluations / mo90 days10,000200,000
EnterpriseCustom — coming soonCustomCustomCustomConfigurableCustomCustom

Overage: Pro +$49 per extra 500k events; Business +$99 per extra 2M events.

What each plan includes

  • Free — everything you need to ship a real experiment: variants and allocation, latency, cost, and business/conversion metrics, FDR correction, the SRM hard gate, live histograms, all three SDKs (Python, TypeScript, Java), and unlimited experiments. One seat.
  • Pro — everything in Free, plus LLM-as-judge quality grading (built-in metrics and custom judge metrics, 50k evaluations a month), 10 seats, higher ceilings, event overage, and email support.
  • Business — everything in Pro, plus 500k judge evaluations a month, 50 seats, method overrides, higher statistical ceilings, 90-day retention, and priority support.
  • Enterprisecoming soon: custom limits, SSO, a security review, and dedicated support. Register interest and we will contact you when it opens.

Why LLM-as-judge starts at Pro

Judge grading reads a sample of your production outputs with an LLM and scores them against a rubric — every evaluation is a real model-inference call with real cost. The free tier therefore includes the metrics the SDK can capture for free (latency, token cost, and business events), and judge-scored quality metrics unlock at Pro with a monthly evaluation allowance. Evaluations are sampled, cached, and hard-budgeted, so grading never runs away with the allowance.

Seats & inviting your team

A plan's seats are per organization: Free 1, Pro 10, Business 50. Invite teammates by email from the dashboard (Invite team in the sidebar); everyone in the org shares the same experiments, metrics, and results. When the org is at its seat cap, the invite flow points at the upgrade path instead of failing silently.

What counts as an event

Your monthly allowance counts every event you track: exposure, model_call, and business_event records. assign calls are free — they are pure local compute in the SDK and never hit the API. So the cost of an experiment scales with how much you evaluate, not how many times you bucket a user.

Retention is not your analysis window

A common question: does a 7-day retention mean I can only analyze 7 days? No. Retention only bounds how long raw events are stored; an experiment always analyzes its full duration. See analysis windows & retention for the details. A longer-running experiment on a short-retention plan analyzes the retained tail — upgrading extends the retained window. Finalized results are snapshotted and never deleted.

Statistical ceilings

max_permutations and max_sample_per_arm cap the per-experiment compute cost of the permutation tests and bootstrap CIs. You can always set lower values per experiment; the plan sets the ceiling. Higher tiers raise the ceiling for very large or very high-precision experiments.

Billing & changing plans

Manage your plan, payment method, and usage from the dashboard. Overages are billed per the plan's overage rate. Downgrades take effect at the next billing cycle and never delete your experiments — only the plan limits change. For volume or annual pricing, contact us.

Choosing a plan

Start on Free and validate your first experiment end to end with latency, cost, and business metrics. Move to Pro when you want automatic quality grading, teammates, or more than 50k events a month. Choose Business for production-scale volume and longer retention. Register interest in Enterprise if you need negotiated limits, SSO, a security review, or a dedicated contact.