AI Summary
About
Patronus AI is an evaluation, monitoring, and guardrail platform for large language models and AI agents. Its products — surfaced on the pricing page as Patronus Experiments, Logs, Traces, Comparisons, and Datasets, plus an evaluator API and the Percival product — let teams test, score, and observe LLM and agent behavior in development and production.
The company sells to two clearly separated audiences. Individual developers and smaller teams self-serve onto a free Developer plan with no credit card, while larger organizations that need unlimited usage, on-prem or dedicated-VPC deployment, SSO, and custom data retention engage sales through a “Book a call” Enterprise motion. The site carries AICPA SOC, TISAX, and HIPAA-compliant trust badges, reinforcing the enterprise-readiness positioning.
Patronus AI was founded in 2023 by ex-Meta machine-learning researchers Anand Kannappan and Rebecca Qian, and launched out of stealth in September 2023 with a $3M seed round led by Lightspeed Venture Partners (TechCrunch, 2023-09-14). In May 2024 it raised a $17M Series A led by Notable Capital (formerly GGV Capital), with participation from Lightspeed and Datadog, bringing total funding to roughly $20M (TechCrunch, 2024-05-22). The company is best known publicly for its open-source Lynx hallucination-detection model (July 2024) and the Percival agent-debugging product (May 2025) — research and product launches that drive its brand far more than its pricing page does. ARR, valuation, and headcount are not publicly disclosed.
Pricing summary : free Developer plan plus pay-as-you-go evaluator API
Patronus AI uses a freemium model with an optional pay-as-you-go API, and a separately quoted Enterprise tier. There is no published fixed-price subscription tier on the live pricing page. The pricing has three dimensions:
- Developer plan (free): $0, no credit card required. Experiments, Logs, and Traces are capped to the last 2 weeks of data; 2 projects and 5 experiments per project; unlimited Comparisons and Datasets.
- Patronus API (optional, metered): $10 per 1,000 small evaluator API calls, $20 per 1,000 large evaluator API calls, and $10 per 1,000 eval explanations. Accounts start with $10 in free credits.
- Enterprise (quoted): “Contact us” / “Book a call”. Everything unlimited, on-prem / dedicated VPC, SSO, custom data retention, higher rate limits, volume discounts, and AI services.
What makes this different: the free plan gates on data retention (a rolling 2-week window) rather than feature access, while the only metered dimension lives in an optional API where evaluator “size” (small vs large) — not raw token count — is the price lever. It is a textbook freemium-pricing base with an opt-in metered layer on top.
Pricing by product
Patronus platform (self-serve plan)
| Tier | Price | Included | Key mechanics |
|---|---|---|---|
| Developer | Free | Experiments (last 2 weeks, 2 projects, 5 experiments/project); Logs & Traces (last 2 weeks); unlimited Comparisons & Datasets | No credit card required; data access capped at 2 weeks |
Patronus platform (Enterprise plan)
| Tier | Price | Included | Key mechanics |
|---|---|---|---|
| Enterprise | Contact us | Everything unlimited; on-prem / dedicated VPC, custom data retention, SSO; Patronus Evaluation Runs, webhooks; higher rate limits, volume discounts; custom eval-model fine tuning & eval dataset generation | Sales-led; booked via Calendly “Book a call” |
Patronus API (optional pay-as-you-go)
| Unit | Price | Notes |
|---|---|---|
| Small evaluator API calls | $10 / 1k | Starts with $10 in free credits |
| Large evaluator API calls | $20 / 1k | 2× the small-evaluator rate |
| Eval explanations | $10 / 1k | Billed separately from evaluator calls |
Sales motions across products: PLG / self-serve for the free Developer plan and the optional pay-as-you-go API; sales-led for Enterprise.
Hidden costs : where the metered API bill grows
The Developer plan is free, so the only line items come from the optional Patronus API. The numbers below are built directly from the published per-1k rates ($10/1k small evaluator calls, $20/1k large evaluator calls, $10/1k eval explanations) and the one-time $10 credit.
A team running 50,000 large-evaluator calls a month with explanations
| Line item | Monthly cost |
|---|---|
| 50,000 large evaluator API calls (50 × $20/1k) | $1,000 |
| 50,000 eval explanations (50 × $10/1k) | $500 |
| First-month $10 credit (applied once) | −$10 |
| Total (first month) | $1,490 |
Because explanations are billed as a separate meter, turning them on effectively adds 50% to the cost of every large-evaluator run at these rates — the explanation line is the easiest cost to overlook.
Want to estimate your own Patronus AI bill? Use the Patronus AI pricing calculator to model your monthly cost based on evaluator call volume.
Pricing evolution : a placeholder template, then ~18 months of a stable freemium + metered API
Wayback snapshots of patronus.ai/pricing tell a tidier story than the company’s research-heavy press would suggest: a single repricing from an unfinished 2023 template to a real freemium model, then near-total stability. The widely-noted “Base $25/month” tier was never a shipped price — it lived only in a Lorem-ipsum placeholder page.
Cadence
| Quarter | Price changes | Product / SKU additions | Notes |
|---|---|---|---|
| 2024 Q4 | 1 | 1 | First real pricing replaces the 2023 placeholder: free Developer plan + optional Patronus API ($10/$20/$10 per 1k) + Custom/Enterprise. |
| 2025 Q2 | 0 | 1 | Developer plan formalized with explicit 2-week retention and 2-project/5-experiment caps; Patronus Traces added; Enterprise CTA becomes “Book a call”. API rates unchanged. |
Tracked range: 2023-Q3 through 2026-Q2. Only quarters with visible activity are listed; intervening snapshots (2025-Q3 through 2026-Q2) show no pricing change.
Notable changes
- 2023-09 — Earliest snapshot is a placeholder template: an Individual (Free) / Base ($25/Month) / Enterprise table priced in generic “pages” with “Lorem ipsum” copy. Not a real product price. (Wayback)
- 2024-12 — Live freemium pricing appears: free Developer plan, optional pay-as-you-go Patronus API ($10/1k small, $20/1k large, $10/1k explanations, $10 free credits), and Custom/Enterprise “Contact Us”. (Wayback)
- 2025-04 — Developer plan refined to its current form (2-week retention, project caps, Patronus Traces added); Enterprise CTA standardized to “Book a call”. (Wayback)
- 2024-12 → 2026-06 — Per-1k API rates held flat at $10/$20/$10 across every snapshot in the window: roughly 18 months without a price move.
Community signal on Patronus AI is research- and funding-driven rather than pricing-driven. The most-discussed items are the $17M Series A (May 2024) and the open-source Lynx model (July 2024); Hacker News and Reddit threads on Patronus pricing specifically are minimal (the top HN story on the company peaks at single-digit points), and no pricing-related backlash or walk-back was found.
What’s unique : retention-gated free tier, evaluator-size metering
1. The free plan gates on data retention, not features. Every product — Experiments, Logs, Traces, Comparisons, Datasets — is available on the free Developer plan, but Experiments, Logs, and Traces only expose the last 2 weeks of data. The wall is a rolling window, not a locked feature.
2. The price lever is evaluator size, not tokens. The optional API charges twice as much for large evaluator calls ($20/1k) as for small ones ($10/1k), passing model cost through as a clean 2× multiple rather than per-token metering — a notably simpler credit-and-usage abstraction than most LLM tooling.
3. Metering is opt-in and rides on top of a free plan. The Patronus API is explicitly labeled “optional,” so usage-based charges only begin for users who choose to call it — pure freemium below, pay-as-you-go above.
Strengths & weaknesses
| Strengths | Weaknesses |
|---|---|
| Genuinely free entry with $10 API credits and no credit card | No published fixed-price tier between free and Enterprise |
| Simple per-1k API metering that is easy to forecast | 2-week data-retention wall on the free plan limits longitudinal work |
| Evaluator-size pricing (large = 2× small) exposes model cost as a clean multiple, not opaque per-token math | Enterprise is fully quoted — no “starting at” anchor or example deal published |
| ~18 months of flat API rates ($10/$20/$10) signal pricing stability for budgeting | ”Optional” API metering is easy to under-budget once explanations are turned on (+50% per large run) |
| Strong enterprise-trust posture (AICPA SOC, TISAX, HIPAA badges) backs the sales-led tier | Free tier’s retention limits are buried in feature rows, not surfaced as an upgrade trigger |
Billing UX : free credits, per-1k metering, Calendly enterprise intake
- $10 free API credits — every new account starts with $10 in credits, enough for 1,000 small evaluator calls before any charge.
- Per-1,000-call metering — the optional Patronus API bills in 1k-call units, with distinct rates for small evaluators ($10/1k), large evaluators ($20/1k), and eval explanations ($10/1k).
- “Get started” self-serve onboarding — the Developer plan activates with no credit card required.
- “Book a call” Calendly intake — Enterprise pricing routes to a 15-minute Calendly “Introduction Call” rather than a published price.
- Rolling 2-week data window — Experiments, Logs, and Traces expose only the last 2 weeks of data on the free plan, an implicit usage/retention control.
Strategic wins : why the free-plus-metered structure works
1. A genuinely free, no-credit-card Developer plan as the on-ramp
Patronus AI lets developers self-serve onto the platform with no card and $10 of API credits, removing the friction that usually stalls evaluation-tool adoption. This is a classic product-led growth motion: land the developer first, monetize the API and Enterprise tier later.
2. Metering on evaluator “size” rather than raw tokens
The optional API charges $20/1k for large evaluator calls versus $10/1k for small ones, exposing model cost as a simple 2× lever instead of opaque per-token math. This keeps the bill forecastable, a recurring theme in usage-based pricing design.
3. Clean separation of self-serve and sales-led motions
The page funnels individuals to “Get started” and enterprises to a Calendly “Book a call”, avoiding the trap of a muddled middle tier. The split keeps the published surface simple while preserving room for custom enterprise quoting.
Areas to improve : gaps in the published pricing
1. No published tier between free and Enterprise
A team that outgrows the free Developer plan but is not ready for a sales call has nowhere to land. A self-serve Team plan with longer data retention and committed API credits would fill the gap — a pattern most hybrid-pricing peers already run.
2. The 2-week data-retention wall is easy to miss
Retention limits on Experiments, Logs, and Traces are buried in the feature rows rather than called out as a plan boundary. Surfacing retention as an explicit upgrade trigger would set expectations and reduce churn surprise, in line with transparent usage communication.
3. Enterprise value is described but never bounded
“Everything Unlimited” and “volume discounts” are stated without any anchor price or starting band, forcing every prospect into a call. Publishing a “starting at” figure or example deal would shorten the enterprise buying cycle.
Key takeaways
- Free can gate on retention, not features. Patronus AI keeps every product available on the free plan but caps historical data to 2 weeks — a softer wall that still creates upgrade pressure.
- Expose model cost as a simple multiple. Large evaluator calls cost exactly 2× small ones, making the only usage lever trivial to reason about.
- Seed usage with credits, not trials. $10 in free API credits converts intent into real calls without a time-boxed trial clock.
- Keep the published surface to two doors. A free self-serve plan and a sales-led Enterprise tier avoid an ambiguous middle that confuses buyers.
- A missing mid-tier is a visible gap. With nothing between free and Enterprise, growing teams may stall — a reminder to map the full adoption ladder.
UBP implications
- Metered APIs can ride on top of free plans. Patronus AI shows usage-based billing layered onto a freemium base, so the meter only starts for users who opt into the API.
- Unit choice can encode model economics. Pricing per 1k evaluator calls — split by evaluator size — passes through model cost without exposing tokens, a simpler abstraction than per-token metering.
- Free credits are a usage-based onboarding primitive. Granting $10 of credits aligns the first experience with the metered model rather than a flat trial, smoothing the transition to paid usage.
Sources
- Patronus AI pricing page — independent raw-HTML re-fetch confirmed the $10/$20/$10 per-1k API rates and that the only $25/”pages” content is leftover 2023 template markup, not a rendered plan (accessed 2026-06-04)
- Patronus AI book-a-call (Enterprise intake) (accessed 2026-06-04)
- Patronus AI blog (accessed 2026-06-04)
Bottom line
Patronus AI keeps its published pricing deliberately simple: a free, retention-gated Developer plan, an optional pay-as-you-go evaluator API metered per 1,000 calls, and a quoted Enterprise tier for unlimited, secured deployments.
Want to compare Patronus AI against other LLM-evaluation and observability pricing? Browse the pricing blueprint.
Pricing timeline : Major events on a vertical axis
Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.
Structure stable: free Developer + pay-as-you-go API + quoted Enterprise
Live pricing page matches the April-2025 structure unchanged — free Developer plan (2-week retention), optional Patronus API ($10/$20/$10 per 1k, $10 free credits), and a quoted Enterprise via 'Book a call'. API rates have held since at least December 2024. Source: https://patronus.ai/pricing
Developer plan adds explicit 2-week retention, project caps, and Patronus Traces
Developer plan is restructured into the current shape: explicit 'Access to last 2 weeks' limits on Experiments, Logs, and the newly added Patronus Traces; 2 Projects and 5 Experiments per Project; Comparisons and Datasets unlimited. Enterprise CTA becomes 'Book a call'. API rates unchanged ($10/$20/$10 per 1k). Source: https://web.archive.org/web/20250403141740/https://patronus.ai/pricing
First real pricing: free Developer + optional Patronus API + Custom/Enterprise
By this snapshot the page shows the live freemium model: a free Developer plan (Patronus Logs, Experiments, Datasets, Comparisons, with some features in 'Preview Mode'), an optional pay-as-you-go Patronus API at $10/1k small evaluator calls, $20/1k large, $10/1k eval explanations ($10 free credits), and a Custom/Enterprise 'Contact Us' tier. No fixed-price mid-tier. Source: https://web.archive.org/web/20241208012731/https://patronus.ai/pricing
Placeholder template pricing (Individual / Base $25 / Enterprise, in 'pages')
Earliest Wayback snapshot of patronus.ai/pricing is an unfinished template: 'Lorem ipsum' subtitle, an Individual (Free) / Base ($25/Month) / Enterprise table priced in generic 'pages' (20/600/Unlimited). The $25 'Base' tier was template scaffolding, not a shipped product price. Source: https://web.archive.org/web/20230919054520/https://patronus.ai/pricing
- · Patronus AI's very first pricing page (Wayback, Sept 2023) was an unfinished template: it listed an 'Individual / Base $25-a-month / Enterprise' table priced in generic 'pages' with literal 'Lorem ipsum' body copy — the $25 tier never reflected a real product.
- · Patronus AI's Developer plan is fully free with no credit card required, but historical data access (Experiments, Logs, Traces) is capped to the last 2 weeks — a soft retention wall rather than a feature gate.
- · The optional Patronus API is metered per 1,000 calls, and 'large evaluator' calls cost twice as much ($20/1k) as 'small evaluator' calls ($10/1k) — the only place model size shows up directly in the price.
Questions & answers
- Is Patronus AI free?
- Yes. The Developer plan is free with no credit card required, though historical data access for Experiments, Logs, and Traces is limited to the last 2 weeks. New accounts also get $10 in free API credits.
- How much does the Patronus API cost?
- The optional API is metered per 1,000 calls: $10 per 1k small evaluator API calls, $20 per 1k large evaluator API calls, and $10 per 1k eval explanations.
- How is Patronus AI Enterprise priced?
- Enterprise pricing is not published. Buyers book an introduction call; the plan adds unlimited usage, on-prem / dedicated VPC, SSO, custom data retention, higher rate limits, volume discounts, and AI services.
- What are the limits on the free Developer plan?
- Patronus Experiments are capped at 2 projects and 5 experiments per project with 2 weeks of access; Logs and Traces also keep 2 weeks of data; Comparisons and Datasets are unlimited.