Vercel AI products are the services Vercel offers beyond web hosting for building and running AI applications: AI Gateway, v0, Vercel Sandbox, Workflows and Vercel Agent. The AI Product Terms define “AI Products and Services” as “the suite of artificial intelligence products and services with artificial intelligence capability, offered by Vercel, including v0 and AI Gateway”.[1] Unlike seats on the core platform, these products are mostly licensed through prepaid credits or per-unit metering, and some have their own plan structure.
Editions
| Product | Plan structure | Billing basis |
|---|---|---|
| AI Gateway | Free tier for every team; paid tier after purchasing credits[2] | Provider list price per token, no markup[2] |
| v0 | Free, Premium (being sunset), Plus, Business, Enterprise[3] | Per-user plan fee plus credits[3] |
| Sandbox | Hobby allowance; Pro and Enterprise pay metered rates[4] | Eight metered dimensions[4] |
| Workflows | Hobby allowance; Pro on-demand[5] | Events, data written and data retained[5] |
| Vercel Agent | Charged per workload[6] | Provider rate plus $0.25 per million tokens[6] |
The AI Product Terms licence
The AI Product Terms grant “a non-exclusive, non-transferable, revocable, limited license (without the right to sublicense), solely for the Term” for personal or internal business use.[1] Use is associated with the customer’s Vercel account, which must be in good standing throughout the term.[1] Customers must not include sensitive personal information such as health information, financial account numbers or government identifiers in their Input.[1] The Acceptable Use Policy adds restrictions for AI Services, including a bar on building competing products or training competing models and on buying, selling or transferring Vercel’s third-party AI providers’ API keys.[7] Catalog: AI Products licence is revocable, limited and for the Term.
Metrics
AI Gateway tokens and credits
“AI Gateway charges no markup and no platform fee on tokens.”[2] A customer buys AI Gateway Credits and charges are deducted from the balance; the charge per request depends on the provider and model and on the input and output tokens processed.[2] Customers are responsible for any payment processing fees, while Enterprise teams can pay by invoice, which has no processing fees.[2] Budgets can cap spend by team, project, API key or team member.[2] Catalog: AI Gateway token; AI Gateway tokens carry no markup.
Free and paid tiers. Every team gets a free tier covering a subset of models, with lower per-model rate limits and a 429 error when a limit is exceeded. Purchasing credits moves the team to the paid tier, raises rate limits and ends the monthly free credit.[2] Bring-your-own-key is available on the paid tier with no markup, but if a request with the customer’s credentials fails, AI Gateway retries with system credentials and charges that fallback usage against the credits balance.[2] Catalog: AI Gateway free and paid tiers.
Add-on surcharges. Some capabilities are off by default and carry charges beyond token rates, deducted from credits.[2]
| Capability | Charge |
|---|---|
| Custom Reporting write | $0.075 per 1,000 tag, user ID or quota entity ID writes |
| Custom Reporting query | $5 per 1,000 queries |
| Team-wide provider allowlist (Pro and Enterprise) | $0.10 per 1,000 successful requests |
| Team-wide zero data retention (Pro and Enterprise) | $0.10 per 1,000 requests |
The tables above are taken from the vendor documentation.[2] Per-request provider filtering costs nothing, and per-request zero data retention carries no additional cost on Pro and Enterprise.[2] Trace Drains are billed separately through Drains usage at $0.05 per 1,000 traces and $0.50 per GB of trace egress, with no allowance on Pro.[2]
Credit terms. Under the AI Product Terms, credits “expire one (1) year from the date issued and do not roll over subsequent periods”, are not redeemable for other Vercel products, are not legal tender and are not refundable.[1] Payments are final and non-refundable unless Vercel authorises otherwise in writing.[1] Catalog: AI credits expire after one year and are non-refundable.
v0 plans and credits
v0 has five plans.[3]
| Plan | Price | Included credit |
|---|---|---|
| Free | $0 per month | Current offer (new signups receive a one-time $5 per user) |
| Premium | $20 per month | $20 per month; being sunset and closed to new users |
| Plus | $30 per user per month | $30 per user per month |
| Business | $100 per user per month | $30 per user per month |
| Enterprise | Custom | Custom |
The tables above are taken from the vendor documentation.[3] Plus, Business and Enterprise have a shared credit pool and centralized billing; data opt-out starts at Business; SAML SSO and priority access are Enterprise only.[3] Monthly credits roll over and expire after 65 days; purchased credits remain after a subscription ends but need an active paid plan to use and expire one year after purchase; shared pool credits expire one year after purchase.[3] When credits run out, generation pauses.[3] Catalog: v0 credit; v0 Plus plan; v0 Business plan; v0 monthly credits expire after 65 days.
The AI Product Terms add that publicly posted v0 content may be viewed and copied by others, and that an Enterprise or opt-out-entitled subscription carries a warranty that v0 Customer Content is not used to train models without consent.[1] A ChatGPT Plus or Pro subscription can cover eligible responses from supported OpenAI models, though image generation and delegated subagents may still use v0 credits.[3]
Vercel Sandbox
Sandbox usage is metered on Active CPU, Provisioned Memory, creations, data transfer, snapshot storage, drive storage, reads and writes.[4]
| Dimension | Hobby included | Pro and Enterprise |
|---|---|---|
| Active CPU | 5 hours per month | $0.128 per hour |
| Provisioned Memory | 420 GB-hours per month | $0.0212 per GB-hour |
| Creations | 5,000 per month | $0.60 per million |
| Data transfer | 20 GB per month | Pro: in Flat Rate CDN; Enterprise: $0.15 per GB |
| Snapshot storage | 15 GB lifetime | $0.08 per GB-month |
| Concurrent sandboxes | 10 | 10,000 |
| Maximum session | 45 minutes | 24 hours |
The tables above are taken from the vendor documentation.[4] Rates apply to the default iad1 region and vary by region for CPU, memory, data transfer and drive.[4] Hobby teams are not charged for extra use, but sandbox creation is paused until 30 days after first use of the feature; Pro usage is charged against the $20 monthly credit.[4] Sandbox failover regions are excluded from Pro trials.[8] Catalog: Sandbox Active CPU; Hobby Sandbox pauses at quota.
Workflows
Workflows are billed on three resources: Workflow Events, Workflow Data Written and Workflow Data Retained.[5] Hobby includes 50,000 events per month and 1 GB written; on demand, events cost $0.02 per 1,000, data written $0.50 per GB and data retained $0.50 per GB-month.[5] “Every state transition in a workflow run is persisted as an event”, and a normal step produces three events with each retry adding one.[5] Retention after run completion is 1 day on Hobby, 7 on Pro and 30 on Enterprise. Functions invoked by workflows and the Vercel Queues they use are charged at their normal rates.[5] Catalog: Every workflow state transition is a billable event.
Vercel Agent
Vercel Agent “charges the underlying provider inference rate with no markup, plus a Vercel Token Rate of $0.25 per million tokens”.[6] Chat and Slack have a limited number of simple requests included during beta; Code Review has no included usage; Observability Plus includes 10 investigations per billing cycle; installation is free.[6] The Vercel Token Rate applies to input, output and cached tokens.[6] Catalog: Vercel Agent adds a $0.25 per million token rate.
Counting and floors
- No MIU payment. For Enterprise or reseller-billed customers, AI fees must be paid by a valid credit card and “are not eligible to be paid via pre-committed Managed Infrastructure Units (MIUs) or credited against other payments you make to Vercel”.[1] Catalog: AI fees cannot be paid from Enterprise committed units.
- Plan changes. An upgrade to a v0 subscription takes immediate effect with full charge for the new period; a downgrade takes effect when the current plan expires.[1]
- Fee changes. Vercel may modify fees, features, payment terms or availability of a Subscription Plan in its sole discretion, effective at the next period.[1]
- Spend caps. Pausing production deployments through Spend Management does not stop AI Gateway key usage or v0 usage billed to the team.[9]
Virtualization and partitioning
Not applicable. Customers are responsible for any application they offer that calls AI Gateway, and anyone interacting with AI Gateway through that application must comply with the AI Product Terms and the terms of any third-party AI provider.[1]
Cloud and BYOL
BYOK for AI providers is the only bring-your-own element, and it is available on the paid AI Gateway tier.[2] Third-party AI providers available through AI Gateway are Non-Vercel Services governed by the provider’s terms.[1] Customers who enable third-party tools or agents to access the Services are bound by their actions and responsible for costs they incur.[10] Catalog: Customer bears costs incurred by third-party tools and agents.
Programs
- AI Gateway free tier. A subset of models at lower rate limits; credits start with the first request.[2]
- Stealth Models. Models released only for test purposes are Beta Features, carry assumed risk, and may have providers who use inputs to train their products.[1]
- Enterprise training warranty. Enterprise AI Gateway subscribers receive a warranty that content is not used to train or improve the AI Products.[1] Catalog: Enterprise AI Gateway content is not used for training.
- Observability Plus investigations. Include 10 Vercel Agent investigations per billing cycle.[11]
Audits and compliance
The AI Product Terms contain no audit right over customer use; compliance is enforced through metering, credits and suspension.[1] The Terms of Service say that on Hobby and trial Pro plans Vercel may use customer content to train AI models, while paid Pro is opt-in and Enterprise can opt out.[10] Customers remain responsible for evaluating AI output, including human review where legally required.[1]
Out of scope
- Open source software that Vercel publishes, which the Terms of Service exclude from the Services.[10]
- Third-party AI provider model terms and prices, which sit with the provider.
- Enterprise negotiated discounts on AI Gateway token spend, which are described as available but are not public.[2]
- Core platform plans, covered in Vercel Pro plan, seats and monthly credit.