Pricing
Simple pricing for one API to every model.
One plan for the platform, tokens at cost
The plan covers routing, policies, guardrails and logs on every model and key. Managed tokens are billed at the provider price plus a flat 6%; on your own keys your provider bills them.
Free
Build on managed models within daily limits. No card required.
- 5,000 requests a month, on any model and any key
- Managed models up to $2.00 per 1M output tokens, for chat, embeddings and rerank
- 100 managed requests and $0.50 of model usage a day, free
- Your own provider keys and cloud accounts
- Smart routing, policies, guardrails and logs
- OpenAI-compatible API
- Community support
Pro
For applications in production, on our models, your keys or your cloud.
Everything in Free, plus
- 50,000 requests a month, then $5 per 25,000 more
- No daily limits on managed inference
- Images, speech, transcription and video on managed inference
- Managed tokens at the provider price plus a flat 6%, from credit you add
- Rate limits up to 600 requests per minute per key
- Spend caps, budgets and alerts
Self-serve upgrades open soon. Start on Free and join the waitlist under Billing.
Enterprise
For teams adopting at scale, on terms and controls agreed by contract.
Everything in Pro, plus
- Volume and committed-use pricing
- Payment terms agreed by contract
- Custom model and creator onboarding
- Security review, DPA and procurement support
- Dedicated support and onboarding
- An uptime SLA and dedicated capacity, by contract
- SSO, planned
The plan is the Inferbase charge. It covers the platform on every request, routed or pinned, on any model and any key, and includes a number of requests each month.
For the models, two ways to run them. Bring your own provider: connect your provider keys or cloud account and the tokens are billed by your provider, with nothing added by us. Or use managed inference: Inferbase serves the model on its own provider accounts, free within the free plan's limits, and on Pro per token at the provider price plus 6%, from credit you add.
Shipping LLM features to customers? Design partners run Inferbase on a real workload with Pro at no charge, inference credits and direct support. Apply to the program
One managed request, itemized
What a request on managed inference costs on Pro, line by line: the tokens, a guardrail check when your policy runs one, and the markup. An example in round numbers; every receipt prints the live ones.
On managed inference a request is billed for the tokens the chosen provider served, at its price, plus a flat markup. A guardrail policy can add a check. Every line is itemized on each request's receipt and in the monthly usage. Routing itself is part of the plan, never a line of its own.
Requests
Each successful request counts one toward the plan: 5,000 a month on Free, 50,000 on Pro. Routed or pinned, managed or your own key, it counts the same.
Tokens
The provider's per-token price plus a flat 6%, read live from the provider and printed on every receipt. No minimum, no idle charge.
Guardrails
Off unless your policy turns them on. An unsafe-content check is a guard-model call, billed per token plus the same 6%. The injection check is free.
Your own provider
On your own keys or cloud account the tokens are billed by your provider. Inferbase charges any guardrail check, and the request counts toward the plan.
- Tokens in1,000 at $0.20 per 1M, the provider's price
- $0.0002
- Tokens out500 at $0.60 per 1M, the provider's price
- $0.0003
- Guardrail checkonly if your policy turns one on: 250 guard-model tokens at $0.20 per 1M
- $0.00005
- Inferbasea flat 6% on the tokens and the check
- $0.000033
- Total
- $0.000583
The rates are examples; every receipt prints the live ones. On your own provider key the token lines are billed by the provider; a guardrail check is still charged, because the guard model runs on ours. Routed or pinned, each request counts one toward your plan's monthly requests.
Compare plans
Every control ships on every plan: pools, presets, failover, guardrails, the decision audit, budgets. The plans differ on price, included requests, limits and support.
| Features | Free | Pro | Enterprise |
|---|---|---|---|
| Plan | |||
| Price | $0 | $19 / mo | Custom |
| Requests included, every model and key | 5,000 / mo | 50,000 / mo | Custom |
| Beyond the included requests | Paused to the 1st | Paused to the 1st until self-serve billing opens, then $5 per 25,000 | By contract |
| Smart routing | |||
| Every plan | ||
| Inference | |||
| Every plan | ||
| Managed inference price | Free within limits | At cost + 6% | Negotiated |
| Managed inference per day | 100 requests, $0.50 | No limit | No limit |
| Tokens per managed request | 8,192 in, 1,024 out | Model limit | Model limit |
| Images, speech and video on managed | |||
| Every plan | ||
| Platform | |||
| Every plan | ||
| Scale and support | |||
| Rate limit per key | 60 rpm | Up to 600 rpm | Custom |
| Managed requests per minute, per organization | 10 | Per key | Custom |
| Support | Community | Community | Dedicated |
| SSO | Planned | ||
| Uptime SLA | By contract | ||
| Dedicated capacity | By contract | ||
| Custom model onboarding | |||
| Compliance support | By contract | ||
Rate limits are per API key and can be set lower on any key. Enterprise terms such as an uptime SLA, dedicated capacity and compliance support are arranged by contract. SSO is planned. Contact us for higher limits or a custom plan.
Frequently asked questions
How the plans are priced, what counts as a request, the two ways to run models, and what is arranged by contract.
Two things are priced. The plan pays for the platform: routing, policies, guardrails, logs and the API, on every model and every key. Free includes 5,000 requests a month, Pro is $19 a month with 50,000, and Enterprise is agreed by contract. Managed inference, the models Inferbase serves on its own provider accounts, is billed per token at the provider's price plus a flat 6%, from credit you add. On your own provider keys or cloud account, your provider bills the tokens and Inferbase adds nothing to them.
Every request that succeeds, whatever it does: chat, embeddings, rerank, images, speech, transcription or video, on a managed model, your own provider key or your cloud account, routed by Inferbase or pinned to a model you named. A request that fails is not counted. The count is per organization and resets on the 1st of each month at 00:00 UTC.
On Free, requests are refused with a 402 and a message naming the limit until the month resets. On Pro the same applies until self-serve billing opens; after that they keep running, and each further 25,000 requests are $5, billed with the next month. Enterprise volumes are set by contract.
A free plan, not a trial, and no card is required. It includes 5,000 requests a month on any model and any key. On Inferbase's own models it covers chat, embeddings and rerank at no charge on models priced up to $2.00 per 1M output tokens at the provider, within daily limits: 100 requests and $0.50 of model usage a day, 10 requests a minute, and up to 8,192 input and 1,024 output tokens per request. A dearer managed model runs from credit, which Inferbase grants on request until self-serve top-ups open. Images, speech, transcription and video on managed inference need Pro. Pro lifts the daily limits.
Credit pays for managed inference on Pro. When it cannot cover a request, the request is refused before any model is called, with a 402 and a message that names the reason, so nothing runs unpaid and nothing is billed without credit behind it. Requests on your own keys and cloud accounts keep running. A project that reaches its monthly budget is refused the same way until the budget is raised or the month rolls over.
$10. Credit is only used for managed inference, so a team that runs entirely on its own provider keys or cloud accounts never needs to add any.
Per token, as the model provider bills it: prompt tokens and completion tokens at the provider's published price, plus a flat 6%. Each request's tokens and price are on its receipt, and the monthly usage view totals them by key and project. Credit is prepaid; nothing is billed after the fact.
Yes, on every plan, and on cloud accounts (AWS Bedrock, Azure, Google Vertex) too. Requests to those models run on your account and your provider bills the tokens, with nothing added by Inferbase. They count toward your plan's monthly requests like any other, and a guardrail check is charged if your policy runs one, because the guard model runs on ours.
Limits are per API key, plus one per organization on the free plan's managed models. On Free, a key allows 60 requests per minute, and managed inference allows 10 requests a minute per organization. On Pro, a key allows 600, and a key that was at 60 is raised when the plan starts; any key can be set lower in its settings. Enterprise limits are set by contract. A request over a limit is refused with a 429 and can be retried.
Yes, on every plan. Organizations hold projects, projects hold keys and a monthly budget, and members carry roles (owner, admin, billing, member, viewer) that gate what they can change and see. Enterprise terms such as an uptime SLA, dedicated capacity and compliance support are arranged by contract; contact us and we will work out the details. SSO is planned and not available yet.
Put intelligence in the middle.
One gateway in front of every model, with your policies applied and every decision on record. Start on the free plan with 5,000 requests a month, no card required.