Promptfoo Pricing in 2026 - What's Actually Free and When You Pay

I get asked “what does Promptfoo cost” a lot, and the honest answer surprises people. For most teams, it costs nothing - the free tier is the real product, not a crippled trial. Promptfoo is MIT-licensed and free forever, with all core eval and red-teaming included. The catch is not a paywall on features. It is a single usage cap, and a wall of “contact sales” above it. Let me decode both.

The pricing, decoded

Here is the whole picture.

Tier	Price	What you get	The limit
Community	$0	All core eval, every provider, red-teaming, self-host	10k red-team probes/mo
Enterprise	Contact sales	Dashboards, team controls, monitoring, SSO, compliance, SLA	Custom
On-Premise	Contact sales	Enterprise features in your own network, air-gapped	Custom

The Community edition is not a stripped-down free tier - it is the full CLI. You get side-by-side prompt and model comparison, assertions, LLM-as-judge scoring, and Promptfoo’s signature red-teaming with 50+ attack plugins and OWASP LLM Top 10 presets. It runs local or self-hosted. It is genuinely free forever under MIT. Full facts are on the Promptfoo page.

So where does money enter the picture? Two places.

The one cap that matters - red-team probes

The single hard limit on the free tier is 10k red-team probes per month. A probe is one attack attempt during a security scan. Regular eval work does not burn this budget - it is specific to red-teaming.

Here is how to think about your own usage. A red-team run fires probes across the attack plugins you enable - prompt injection, jailbreaks, PII leakage, and the rest of the 50+ plugins - against each prompt or endpoint you point it at. A single ad-hoc scan on one target barely dents 10k. But a nightly CI scan across many plugins and many prompts adds up fast, because every plugin-times-prompt combination fires its own attempts. If red-teaming is a once-a-sprint activity, you will never see the cap. If it is a gated CI step running on every merge, model it before you assume free covers you. I would not guess your exact probe count from the plugin list - run one scan, read the report, and multiply by your cadence.

The contact-sales gap

The second place money enters is the jump to Enterprise, and this is the real gotcha for buyers. Everything a bigger team needs sits behind “contact sales” with no public price. Centralized dashboards, team management, continuous monitoring, custom attack profiles, compliance frameworks, SSO, SLA support, a dedicated runner - all of it is Enterprise or On-Premise, both contact-sales only.

I could not find a published Enterprise figure, so I am not going to invent one. If you are building a budget, assume a sales call, not a pricing page. That opacity is the honest downside of an otherwise generous model - the free tier is fantastic, but the moment you need team features, you lose the ability to self-serve a price.

One more thing to weigh, separate from cost. Promptfoo was acquired by OpenAI on 9 March 2026. Both sides state it stays open source under MIT, and existing customers keep being supported. The free tier is not at risk today. But long-term OSS governance under a single large owner is a fair thing to file away before you build a multi-year workflow on it.

Cheaper to scale - the tools with public pricing

Since Promptfoo’s free tier is already the cheapest thing in the category, “cheaper alternative” is the wrong frame. The real question is what you pay for hosted collaboration and dashboards - the stuff Promptfoo gates behind sales. Two tools publish those numbers.

DeepEval is the closest philosophical match, and its cloud pricing is public. The OSS framework is Apache-2.0 and free, like Promptfoo, but with a pytest-style, SDK-first approach instead of YAML config. If you want a hosted UI, Confident AI cloud is public at $200/mo for the Starter tier - 5 GB-month of trace spans and 5 projects. Know the shape before you commit though: there is a steep 10x cliff from Starter to the $2,000/mo Team tier, with nothing in between. For a Python team that wants public pricing and programmatic tests, DeepEval is the natural comparison. The DeepEval page has the full tier breakdown, and I compared the two approaches directly in DeepEval vs Promptfoo.

Braintrust is the pick if evals and CI regression gates are the point, not red-teaming. Its free Starter is $0 with unlimited users, and Pro is a public $249/mo. But the meter is the thing to watch - Braintrust bills “processed data” in GB, counting every byte of inputs, outputs, prompts and metadata, with no hard spending cap. Verbose agents and big RAG contexts burn that allowance fastest. Set billing alerts on day one. I dug into the meter in the Braintrust pricing breakdown.

So what should you actually pay?

• Solo dev or small team doing eval and occasional red-teaming - pay nothing. Promptfoo Community is the real product, and 10k probes a month covers a lot.
• You need team dashboards, SSO or compliance - you are in a Promptfoo sales conversation with no public price, so get a quote and compare it against DeepEval’s public $200/mo and Braintrust’s public $249/mo.
• Python team that wants public pricing and pytest-style tests - DeepEval, with the $200-to-$2,000 cliff in view.
• Regression gates over red-teaming - Braintrust, with billing alerts on from day one.

The takeaway is simple. Promptfoo’s free tier is one of the best deals in the category, and the only number you truly need to watch is 10k red-team probes a month. It is the Enterprise pricing, not the free tier, that will cost you a phone call. For a wider view, see the best LLM eval frameworks and Promptfoo alternatives.

All prices here were read from each vendor’s own pages and verified on 26 July 2026. This category ships breaking changes monthly, so we re-verify every 30 days.