Freemium Pricing: Examples & Companies

What is it

Freemium Pricing is a pricing model that combines a permanently free tier with paid upgrade plans, used to drive product-led growth and self-serve acquisition.

The free tier is not charity — it is the top of the funnel. It gets a user into the product at zero friction, demonstrates value, and relies on a minority who hit its limits (or want better models, higher limits, or commercial rights) to upgrade. It is the most common self-serve model in the corpus: 169 companies run a freemium tier.

The canonical shape is a three-rung ladder. OpenAI runs ChatGPT Free → Plus $20/mo → Pro $200/mo; Anthropic runs Claude Free → Pro $20/mo → Max (up to $100/mo); and Google runs Gemini free → AI Pro $19.99/mo → AI Ultra. The same free front door recurs across creative tools (Suno, HeyGen, Ideogram, Runway), search and research (Perplexity AI), and productivity (Notion AI, Gamma, Canva).

What separates freemium from a discounted trial is permanence: the free tier never expires. That changes the design problem from “how do I force a decision at the deadline” to “how do I let a user live on free indefinitely and still convert.” For a deeper primer on how free allowances interact with metered plans, see our introduction to usage-based pricing and the guide to usage-based pricing models.

How it works

Two design decisions define a freemium plan: what the free tier withholds and what triggers the upgrade. Across the corpus, the dominant answer to the first is a volume cap, and the dominant answer to the second is a better model or a higher limit.

A worked example. Take a solo video creator on Runway. The free plan gives a fixed one-time credit grant to try text-to-video; once that runs out, the next rung is Standard at $12/mo (billed annually), then Pro at $28/mo and Unlimited at $76/mo. If the creator generates enough clips to exhaust free credits in a week, the $12/mo tier pays for itself against a single paid stock-video clip — that is the upgrade trigger working as designed: the free tier proved the value, and the volume ceiling forced the decision.

The math that makes this viable is the marginal cost of a free user. For DeepSeek — whose API runs as low as ~$0.14 per 1M input tokens — free inference is cheap enough to subsidise a large free base. For a video model where each generation burns real GPU-seconds, the free grant is a fixed one-time allotment, not a renewing monthly pool, precisely because the marginal cost is high.

Why volume caps beat feature gates: withholding features hides the product’s value and weakens the upgrade case. Capping volume lets the user experience the full product, then upgrade when they outgrow the free allowance — which is why most corpus freemium tiers cap usage, not capability. See the guide to choosing the right usage metric for how to pick the cap that best predicts willingness to pay.

Companies using this

One hundred and sixty-nine companies in the current corpus run a freemium tier, spanning frontier assistants (OpenAI, Anthropic, Google, DeepSeek, Mistral AI, Perplexity AI), creative media (Suno, HeyGen, Ideogram, Recraft, Runway, Descript), productivity (Notion AI, Gamma, Canva, DeepL), and developer tools (GitHub Copilot, Replit AI, Codeium, Exa). The table below lists how each structures the free-to-paid jump.

Patterns observed

• A $100–$200 power tier now sits above the $20 anchor. Free → ~$20 was the 2023 default; in the last year the frontier assistants stacked a power tier on top, turning a binary free-or-$20 choice into a three-rung upgrade path. The mass tier still does the volume; the power tier captures the heaviest users who would otherwise cap out.

• Free tiers cap volume, not features. Across the set the free plan almost always gives real product access with a usage ceiling. Notion AI gives a limited AI-response allowance inside a full workspace, and Canva gives a genuinely usable free design surface with paid tiers unlocking scale. The bet is that exposure converts better than scarcity.

• Freemium for humans and per-token APIs for machines coexist on the same company. Most frontier vendors run freemium for the consumer app and a pay-per-token API for developers — Google’s Gemini Flash starts around $0.075 per 1M tokens, and Mistral AI runs the same split. The free consumer tier acquires humans; the metered API monetises applications.

• Commercial-use rights are a common paid gate. Creative tools frequently keep the free tier for personal use and charge for commercial licensing. Ideogram gates commercial rights behind paid plans — the upgrade trigger here is legal, not technical, and it converts exactly the users (freelancers, agencies) with the highest willingness to pay.

• Seat-based paid rungs sit on top of individual freemium. Tools that start as a personal free product often layer team plans above the individual paid tier: Gamma moves from a free tier to per-seat Plus and Pro plans, and DeepL prices its paid tiers per user (from ~$8.74/user/mo). The free tier lands the individual; seat pricing captures the team they bring.

Counterexamples & variants

The clearest counterexamples are the gated, no-free-tier enterprise plays. Harvey, Jasper and Glean all sell top-down with sales-quoted, seat-based plans and no self-serve free entry — none carries a free tier in the corpus. Their buyer is procurement, not an individual who needs convincing, so a free tier would add cost without accelerating the deal. When the sales motion is enterprise and the contract is annual, the trial or the proof-of-concept replaces the permanent free tier.

The sharpest single counterexample is Midjourney: a mass-market consumer creative tool with no free tier at all. It dropped its free trial in 2023 after abuse and has run subscription-only since. Midjourney proves that even in the category where freemium is most expected — consumer image generation — a strong-enough brand and waitlist demand can monetise from the first minute. Freemium is a choice, not a law of the segment.

A notable variant is freemium-plus-PAYG: a free credit tier for humans paired with a pay-as-you-go API wallet for developers. HeyGen runs two front doors — a free consumer tier and a metered developer API priced per generation. It is still freemium at the top of the human funnel, but the developer path is pure usage from dollar one. The other common variant is the fixed-grant free tier: where the marginal cost of a free user is high, the free allowance is a one-time credit grant rather than a renewing monthly pool, the shape Runway uses for GPU-heavy video generation.

What this means for buyers vs vendors

For buyers

Use the free tier to validate fit, but model the tier you will actually live on — the free allowance is deliberately just short of a daily workflow. Before committing, answer three questions: what is the real upgrade trigger (a model you need, or just a volume cap you will hit)? Does the free grant renew monthly or is it a one-time allotment (as on Runway)? And do you need commercial-use rights, which many creative tools gate behind paid plans? If your usage is spiky, compare the fixed $20-class tier against a metered API on the same vendor — for OpenAI, heavy programmatic use may be cheaper on the per-token API than on a seat. Model both against your real volume with the pricing calculator, and read the introduction to usage-based pricing for how free allowances interact with metered overage.

For vendors

Freemium works when the marginal cost of a free user is low and the upgrade trigger is sharp. For AI products, free inference is a real cost — cap it on volume, and route free users to cheaper or smaller models to protect margin (the DeepSeek and Google API tiers show how cheap inference can get, which is what makes a large free base survivable). Design the free tier to demonstrate value fully rather than hide it behind a feature gate, then make the upgrade reason obvious: a better model, a higher limit, or commercial rights. If you sell to enterprise procurement — the world of Harvey and Glean — freemium is likely the wrong front door, and a trial or sales-led proof-of-concept fits better. And remember Midjourney: if demand is strong enough to monetise from minute one, a free tier may only add cost. Use the guide to choosing the right usage metric to pick the cap that best separates the free base from the buyers.

Explore this theme in the knowledge graph →