Pricing

flinq is prepaid, pay as you go. You hold a EUR credit balance, and usage is metered and drawn down from it in euros. There is no subscription and no monthly reset. Pricing is per model: each request is billed at the rate of the model that serves it, and every model draws from the same balance. The sections below cover the model rates, the credit packs, and how the meter works.

Beta

flinq is in beta, but billing is live. Sign up in the console, verify your email, and the EUR 5 starter credit is applied automatically — no card required to start.

The principle: you pay when the model runs

flinq bills inference, not plumbing. Parsing a file and composing chunks are transforms on the way to the model; they are free or flat-priced. The same text is never billed twice on its way through the extract → chunk → embed pipeline: chunk your file and embed the chunks in one /v1/chunk call with a model, and you pay for those tokens exactly once.

Models and rates

Model	Endpoints	Price per 1M tokens
flinq-pilot-otter	embeddings, search, rerank, chunk (with model)	EUR 0.40
flinq-pilot-wombat	embeddings, search, rerank, chunk (with model)	EUR 0.40

Document endpoint	Unit	Price
/v1/extract	per document	EUR 0.02
/v1/chunk without model (text chunks)	—	free
/v1/chunk with model (vectorized chunks)	per 1M chunk tokens	EUR 0.40

A model’s rate applies uniformly across its endpoints. Tokens are counted server-side and converted to euros at the serving model’s rate, then subtracted from your balance. Usage is measured and stored in EUR rather than raw tokens, which is what lets models with different rates share one balance.

New accounts start with EUR 5 of free credit, applied once at signup. That is enough to embed roughly 12.5 million tokens before you top up.

Credit packs

Top up with a one-time credit pack. Packs are 1:1: pay EUR 10, get EUR 10 of usable credit. The minimum top-up is EUR 10, and credit never expires.

One-time €10 €10 usable credit approx. 25M tokens

Most chosen One-time €25 €25 usable credit approx. 62M tokens

One-time €50 €50 usable credit approx. 125M tokens

One-time €100 €100 usable credit approx. 250M tokens

The token figures above are the buying power of each pack at the standard rate (EUR 0.40 per 1M tokens); they are not an allowance or a quota. Whatever you buy stays in your balance until you spend it, on any model.

How metering works

• One EUR meter. Every endpoint and every model draws down the same balance, so there is no separate budget per endpoint or per model. Usage is stored in euros, not raw tokens.
• Server-side counting. Tokens are counted on the server with the model’s own tokenizer, not by splitting on whitespace, so the count matches what the model actually processes.
• Tokens are converted to EUR. The token count for a request is multiplied by the serving model’s per-token rate (see the table above) and subtracted from your balance.
• Embeddings and search bill input tokens. /v1/embeddings and /v1/search are metered on the tokens in the text you send.
• Rerank uses the Voyage formula. /v1/rerank is billed on query_tokens × n_docs plus the sum of the document tokens in the candidate pool. A query reranked against 50 documents is charged for the query once per document plus all 50 documents.
• Extract bills per document. /v1/extract costs a flat EUR 0.02 per parsed file, whatever its size. The response’s usage still reports the extracted token count, as a planning figure for downstream chunk and embed calls, not as the billed amount.
• Chunk bills only what gets embedded. /v1/chunk without a model returns text chunks and is free. With a model, the produced chunk tokens are billed once at that model’s rate and the chunks come back vectorized; usage reports exactly the billed tokens. Note that context=headers repeats header text into every chunk by design, which raises the embedded token count.
• A zero balance returns 402. When your balance reaches zero, the next billable request returns 402 until you top up. An optional monthly spending cap, if you set one, returns 402 the same way once it is reached.

See Rate limits and errors for the full status-code behavior.

Auto-recharge and spending cap

Both are optional and off by default, managed on the console’s billing page:

• Auto-recharge. Save a card once through a zero-cost checkout, then set a threshold and a top-up amount. When your balance drops below the threshold, flinq automatically charges the chosen credit pack to the saved card, so long runs do not stall at zero. Every automatic top-up sends an email receipt, and you can switch the feature off at any time.
• Monthly spending cap. Set a euro ceiling on how much you spend per calendar month. Once the cap is reached, requests return 402 until the month rolls over or you raise the cap. It is a guardrail against a runaway job, independent of your balance. You are warned by email at 80 percent and when the cap is reached.

Batch

An asynchronous batch endpoint is a documented fast-follow, OpenAI-shaped and billed at a 50% discount on the serving model’s standard per-token rate. It is the cost-efficient path for embedding a large corpus offline. See Bulk & batch for the synchronous concurrency recipe you can use today and the batch design.

Billing

At general availability, checkout runs through Paddle as the Merchant of Record. Prices are in EUR, and any applicable VAT is shown at checkout.

flinq serves both private and company accounts. An account with a company name and a valid VAT ID (set at signup or any time later in the console settings) is billed as a business: invoices are issued to the company with its VAT ID, and within the EU the reverse-charge mechanism applies where the VAT ID qualifies. Switching between private and company takes effect for future payments and invoices; documents already issued stay unchanged.

During the beta

Self-serve is live: sign up in the console, verify your email, and start with EUR 5 of free credit. Balance, credit top-ups, auto-recharge and the spending cap are all managed on the console’s billing page.