Updated pricingopen-modelsinference-providers

Together, Fireworks, DeepInfra, Groq and Novita: What Open-Model Hosts Say Besides the Token Price

Short answer. Read on October 28, 2026: Together AI is prepaid, with a $15 minimum first purchase, and says its credits do not currently expire. Fireworks AI is prepaid and charges a 1.5x premium on region-restricted deployments. DeepInfra needs a card or prepayment. Groq invoices monthly, after thresholds from $1 to $1,000. Novita offers an introductory 50% batch discount.

In plain words. These companies run open AI models and charge per token. A token is a short piece of text, about three-quarters of a word. Some want payment first. Others send a bill afterwards. No page read here prints a discount for large volumes.

The price per token is only part of what a host’s pages say about paying. Before the first request, a host may want a card, a prepaid balance or a minimum purchase; it may say whether unused credit expires, when it sends an invoice, and what it charges extra for. Those lines are spread across pricing pages, billing pages and FAQs. This post puts them side by side for five hosts, quoted as each page printed them on October 28, 2026.

The per-token prices themselves are not repeated here. They are on open-model prices by host, with one page per model for those that pass AI Signal’s quality filter, such as GLM-5.3, gpt-oss-120b and Kimi K3. On October 28, 2026 all 7 models that two or more current hosts price there stayed within 1.15× of each other.

What each host’s pages say

Where a host’s page says nothing on a point, the table says “not stated on its page”. That is not the same as “no”: it only means the page read on October 28, 2026 does not say.

HostTopicWhat the page says, quotedPageRead
Together AIPayment“You need a positive credit balance to use the platform.” “If your balance reaches zero, API access is suspended until you add credits.”billing (credits)October 28, 2026
Together AIMinimum purchase“The minimum is $15 for your organization’s first purchase and $5 after that.”billing (credits)October 28, 2026
Together AICredit expiry“No, prepaid balance credits in your Together.ai account do not currently have an expiration date.”billing (credits)October 28, 2026
Together AIFree trial“Together AI does not currently offer free trials.”billing (credits)October 28, 2026
Together AISales or volume“Contact sales”, in the “Reserved” column of its GPU table, not for per-token usepricingOctober 28, 2026
Fireworks AIPayment“Pay per token with pre-paid, usage-based billing.” “Simply add a payment method to buy credits, and usage is then deducted from your balance.”pricingOctober 28, 2026
Fireworks AIWhat is billed“Usage is billed on output, input, and cached tokens.”pricingOctober 28, 2026
Fireworks AIMinimum purchasenot stated on its pagepricingOctober 28, 2026
Fireworks AICredit expirynot stated on its pagepricingOctober 28, 2026
Fireworks AIRegional premium“Region-restricted deployments are priced at a 1.5x premium.”pricingOctober 28, 2026
Fireworks AISales or volume“Contact us for enterprise deployments with faster speeds, lower costs, and higher rate limits.”pricingOctober 28, 2026
DeepInfraPayment“You have to add a card or pre-pay or you won’t be able to use our services.”pricingOctober 28, 2026
DeepInfraInvoices“Pre-paid accounts draw down their balance as they go, and invoiced accounts are billed once at the beginning of the following month.”pricingOctober 28, 2026
DeepInfraMinimum purchase“You only pay per token — no idle GPU time, no minimums, no seat fees.”billingOctober 28, 2026
DeepInfraCredit expirynot stated on its pagepricing, billingOctober 28, 2026
DeepInfraSales or volume“Have questions or need a custom solution?”, followed by a “Contact Sales” linkpricingOctober 28, 2026
GroqPayment“Groq uses a monthly billing cycle, where you receive an invoice in arrears for usage.”billing FAQOctober 28, 2026
GroqInvoices“an invoice is automatically triggered and payment is deducted when your cumulative usage reaches specific thresholds: $1, $10, $100, $500, and $1,000.”billing FAQOctober 28, 2026
GroqMinimum bill“We only bill you once your usage has reached at least $0.50.”billing FAQOctober 28, 2026
GroqCredits“Yes! We occasionally offer promotional credits, such as during hackathons and special events.”billing FAQOctober 28, 2026
GroqCredit expirynot stated on its pagebilling FAQOctober 28, 2026
GroqSales or volumenot stated on its pagebilling FAQOctober 28, 2026
NovitaBatch“Batch inference is available at an introductory 50% discount on input and output tokens for supported models.”pricingOctober 28, 2026
NovitaPayment, minimum, credit expirynot stated on its pagepricingOctober 28, 2026
NovitaSales or volume“Talk to Sales” linkpricingOctober 28, 2026

Prepaid balance or a bill afterwards

Three of the five describe a balance you fill before using the API. Together AI says “You need a positive credit balance to use the platform.” and stops access at zero. Fireworks AI sells credits and deducts usage from them. DeepInfra accepts both ways: “Pre-paid accounts draw down their balance as they go, and invoiced accounts are billed once at the beginning of the following month.”

Groq works the other way round. Its billing FAQ says “Groq uses a monthly billing cycle, where you receive an invoice in arrears for usage.”, with earlier invoices for new accounts at $1, $10, $100, $500 and $1,000 of lifetime usage. It also says “These are one-time thresholds for most customers.” and that customers with a billing address in India have different ones.

Minimums and expiry

Only Together AI names a minimum purchase: $15 the first time, $5 after that. DeepInfra’s billing page says the opposite in its own words: “no minimums, no seat fees”. Groq names a smallest bill, $0.50, which is not a purchase.

On expiry, only Together AI answers: its prepaid credits “do not currently have an expiration date”, and “If any changes to this policy are made in the future, Together.ai will notify customers in advance through official communications.” The other four pages read do not say.

Discounts, premiums and sales contacts

No page read here prints a price for large volumes. What they print instead:

  • Fireworks AI says enterprise deployments come with “lower costs”, and sends those customers to sales; it also charges a premium the other way: “Region-restricted deployments are priced at a 1.5x premium.” AI Signal recorded a related change on the price changes page: since September 1, 2026 Fireworks adds 50% to serverless models it marks as US-only.
  • Novita prints a batch discount and calls it introductory: “Batch inference is available at an introductory 50% discount on input and output tokens for supported models.” The page gives no end date. An introductory discount is a promotion, so it is not a price AI Signal would publish as a list price.
  • Together AI’s “Contact sales” sits in the reserved GPU column of its pricing page, about dedicated hardware rather than per-token use. DeepInfra and Novita show a sales link without saying what it offers.
  • Groq’s billing FAQ has no sales contact; it sends billing questions to its support address.

Novita and Groq: what is and is not here

Novita is not in AI Signal’s pricing table or on the open-model pages: its prices have not been added to the dataset, and nothing here compares them. It appears in this post only through the lines quoted above.

Update, October 29, 2026: Novita AI’s prices are now in the dataset. AI Signal read its pricing page on October 29, 2026 and added five of its offerings to the open-model pages. This post and its quotes are unchanged.

Groq is in the dataset, with prices read from its models page on October 26, 2026. Its pricing page could not be read on October 28, 2026: groq.com/pricing redirected to Groq’s home page, so this post gives no Groq price. The Groq lines above come from its billing FAQ.

How we checked

On October 28, 2026 AI Signal read Together AI’s pricing page and its billing page on credits, Fireworks AI’s pricing page, DeepInfra’s pricing page and billing page, and Novita’s pricing page with the same script the weekly price detector uses, which turns each page into plain text. Groq’s pricing page and billing FAQ were read with a headless browser, because they are built in the browser. Each reading is saved with its date, and a test checks every quote in this post against those saved readings. Where a page says nothing on a point, the post says “not stated on its page”. No token price is restated here, and nothing here says which host to choose.

Frequently asked questions

Do Together AI credits expire?

Together AI's billing page, read on October 28, 2026, says its prepaid balance credits 'do not currently have an expiration date', and that customers will be notified in advance if that policy changes.

What is the minimum purchase on Together AI?

$15 for an organization's first credit purchase and $5 after that, according to Together AI's billing page read on October 28, 2026.

Does Fireworks AI charge more for region-restricted deployments?

Yes: Fireworks AI's pricing page, read on October 28, 2026, says region-restricted deployments 'are priced at a 1.5x premium' and points to its sales team.

How does Groq bill pay-as-you-go usage?

Groq's billing FAQ, read on October 28, 2026, describes a monthly invoice in arrears, with earlier invoices when lifetime usage first reaches $1, $10, $100, $500 and $1,000.

Written with AI assistance from this site's own verified pricing dataset and reviewed by Álvaro Lucero Rufino, who is responsible for what stays published here. Every figure links to the page where it was verified, with the date it was read; if a number here and a number on that page ever disagree, the page is the one to trust. Automated publishing was discontinued in September 2026; posts that no longer met that standard were removed rather than left up.