Hybrid Pricing Model: Examples & Companies

What is it

Hybrid Pricing Model is a pricing model that combines a fixed recurring fee with variable usage-based charges, both meaningful to the bill.

The fixed component is typically a per-user seat fee, a platform access fee, or a prepaid credit pool. The variable component meters consumption on the dimension that drives the vendor’s underlying cost — tokens for LLM products, resolutions for AI support agents, GPU-hours for inference clouds, minutes for voice and video. Hybrid has become the default packaging for AI software because a pure subscription under-prices heavy users while pure usage scares off cautious ones.

It is the most common paid structure in the corpus: 144 companies price on a hybrid basis. Intercom charges $29–$132 per seat for platform access plus $0.99 per Fin AI resolution; GitHub Copilot layers usage-based “AI Credits” on top of its Pro, Pro+, and Business seats; inference clouds like Modal pair a fixed plan with per-second GPU billing. What separates them is where each draws the line between fixed and metered, and how aggressively the meter ramps.

Hybrid is also the endpoint of a broader move we track in subscription-to-usage drift: vendors that tried pure usage have mostly re-added a fixed floor, and vendors that started flat have bolted on a meter.

How it works

Every hybrid model picks two things: what the fixed fee buys, and what unit the meter counts.

The unit math depends on whether the fixed fee includes a usage allowance. Two patterns dominate:

Worked example (Pattern A): a Cursor Pro user at $20/month who consumes $34 of model credits pays max($20, $34) = $34 — the seat fee is effectively a floor, and the overage is the delta. Worked example (Pattern B): an Intercom team on a $99/seat plan with 5 seats that resolves 400 tickets via Fin pays (5 × $99) + (400 × $0.99) = $495 + $396 = $891. Pattern A is honest about variable cost but produces unpredictable bills; Pattern B is more familiar but obscures unit economics.

Companies using this

The 144 companies below span every AI category: coding tools (Cursor, GitHub Copilot, Codeium, Sourcegraph Cody), customer-support automation (Intercom, Decagon, Gladly), voice and video (ElevenLabs, Cartesia, Bland AI, Retell AI), inference and hosting clouds (Modal, RunPod, Replicate, Together AI, Groq), LLM observability (LangSmith, Helicone), and billing infrastructure (Stripe Billing, Orb). The table below lists their structural choices side by side — fixed vs. metered, billing units, and free-tier availability.

Patterns observed

• The fixed component tracks how buyers budget: per-seat in B2B, per-account in prosumer, per-platform in infra. GitHub Copilot ($19–$39/user Business/Enterprise) charges per seat; consumer creative tools like Cartesia ($4 Pro → $239 Scale) charge per account; inference clouds like RunPod charge a platform fee that buys access, not seats.

• The variable component meters the dimension that drives vendor cost. Inference clouds meter GPU-time — Modal at $0.001097/sec for an H100, RunPod at $2.89/hr, Replicate at $0.001525/sec dedicated. Voice and media tools meter what they consume: ElevenLabs bills characters and Bland AI $0.14/min. The meter is whatever the vendor pays upstream, marked up.

• The fixed component is increasingly a credit pool, not a seat. A large share of these companies — Cartesia (20K free credits), Apify (prepaid compute units), Roboflow (unified credit metering) — express the fixed fee as a monthly credit allotment the variable usage draws down. See credit-based billing for that mechanic in depth.

• Free or freemium entry is table stakes. Most of these companies offer a free or freemium tier before the paid hybrid plans — Helicone (10K free requests), Retell AI ($10 free credits), and Modal (Starter $0 + $30 credits) all seed usage before the meter starts. The PLG conversion path is no longer optional in this category.

Counterexamples & variants

Hybrid breaks down when the variable component dominates so heavily that the fixed fee becomes a token gate. The June 2025 Cursor incident is the canonical case: heavy users on the $20 Pro plan burned through the credit pool and hit unexpected $10–$20 daily charges with no in-product warning, effectively making the seat fee irrelevant. Cursor’s response — spend visibility, clearer credit accounting, and refunds — became the playbook for the category and fed the broader shift toward in-product spend dashboards and advance-notice policies.

The opposite failure mode is a hybrid where the variable component is too small to matter — legacy SaaS that meters “API calls” at fractions of a cent against a multi-thousand-dollar contract. The metering is theatre, not pricing; the corpus classifies those as subscription, not hybrid.

The genuine counterexamples to hybrid sit at two poles. On one end are the pure-usage voice and inference plays with no fixed floor at all — Retell AI bills strictly $0.07–$0.31/min with no seat, and Vapi passes STT/LLM/TTS through at cost on top of a per-minute host fee. On the other end are the sales-gated, quote-only enterprise variants: Glean publishes no prices, wrapping per-user Enterprise Flex seats and pooled FlexCredits behind a quote, and outcome-only agents like Sierra and Decagon bill per resolution with no seat component. These edges show the boundary of the model: strip the fixed floor and it becomes pure usage; strip the meter and it becomes flat subscription.

What this means for buyers vs vendors

For buyers

Ask three questions of any hybrid pricing page: what’s the included quota, what’s the overage rate, and where do I see real-time consumption? A hybrid plan with vague included quotas and no in-product spending dashboard is structurally bill-shock-prone — invisible usage until the invoice lands is the failure mode to screen out.

Model your two extremes before you sign. Run the light-user case (do you overpay for a floor you’ll never fill?) and the power-user case (does the meter blow past the seat fee?). For credit-pool models like Cartesia, confirm whether unused credits roll over and what an overage credit actually costs. Our introduction to usage-based pricing walks through the value-metric framework you’ll need to compare offers, and the pricing calculator hub lets you model seat-plus-usage totals against your own volume.

For vendors

Hybrid is the right default for any product where customer usage varies more than about 3x. Below that, a flat per-seat fee suffices. Above it, you’re either subsidizing heavy users (pure per-seat) or leaving easy expansion revenue on the table (pure usage). The winners in this corpus — Modal, Together AI, Vercel — meter the exact unit that drives their COGS, so the metered line item stays margin-positive by construction.

The operational half is where most implementations stall — invoicing the metered component, prorating mid-cycle changes, and reconciling prepaid credits against consumption. Get it wrong and you produce churn surveys full of “the bill was a surprise”; see our guide on usage invoicing and billing cycles for the mechanics that keep a hybrid bill defensible.

Explore this theme in the knowledge graph →