Question 1

What does “0% markup” actually mean?

Accepted Answer

You pay the model provider’s native per-token rates with no per-token markup added on top. The price you see in our Models list is the price you pay — there is no surcharge for routing through InferenceHub.

Question 2

Is there a free trial?

Accepted Answer

Yes. Every new account starts with $1 in credit, and no credit card is required to begin. Point your client at the gateway and start making requests right away. The hosted chat also has a set of Free models (GPT-5.6 Luna, Claude Haiku 4.5, and a handful of open-weight models) that cost $0 to chat with — new accounts start there, and the first payment unlocks the full chat catalog.

Question 3

How does the daily reward work?

Accepted Answer

Make at least one successful API request in a day (UTC) and you can claim a small credit in the portal's Rewards tab: $0.10 of open-weight credit on day one, ramping to $0.25 by day seven of a consecutive-day streak. Daily credits expire after 48 hours, so use them as you go. Every 7-day streak adds a $0.50 bonus spendable on any model — Claude and GPT included — and a 30-day streak bumps that bonus to $2.00 (bonuses last 7 days). It's a fixed, published ladder — the same amounts for everyone, no chance involved. Miss a day and the streak restarts at one; your balance and everything else are untouched.

Question 4

How do I add credit, and which payment methods do you accept?

Accepted Answer

InferenceHub runs on a prepaid balance. You can top up with a card or with cryptocurrency. Usage is metered per request and drawn down from your balance.

Question 5

Do you offer subscription plans?

Accepted Answer

Yes — three sizes: All Access ($10/mo), Plus ($25/mo), and Max ($50/mo). Every tier gets every model: premium Claude and GPT models (Opus, GPT-5.6 Sol) share one burstable monthly premium budget ($50 / $200 / $500 by tier — pooled, so spend it on either vendor), and every other model — open-weight daily drivers like GLM-5.2 and Kimi K2.6, plus lower-priced frontier like Sonnet — draws standard 5-hour/weekly windows that scale with the tier. The open-weight models run right inside Claude Code. Deep Research in the hosted chat is included. Reach any limit and you keep working at pay-as-you-go rates, with no hard stop. Prefer no commitment? Pay-as-you-go works without any plan.

Question 6

What happens when I hit my plan’s usage limit?

Accepted Answer

Nothing breaks. Requests beyond your plan’s included usage simply continue at the same 0%-markup pay-as-you-go rates shown in the Models section.

Question 7

Do I need a credit card or KYC to sign up?

Accepted Answer

No. You can create an account with just an email, and you don’t need to provide your legal name, address, or phone number to use the gateway. A card is only needed if you choose to top up by card.

Question 8

Which models can I use?

Accepted Answer

A single API key gives you access to every model on the platform — Claude (Fable, Opus, Sonnet, Haiku), GPT (including Codex), GLM, DeepSeek, Kimi, MiniMax, Qwen, NVIDIA Nemotron, and more. See the Models section on the home page for the current catalog and rates.

Question 9

Are responses identical to calling the provider directly?

Accepted Answer

Yes. InferenceHub is a byte-faithful gateway: your request and the model’s response pass through unchanged on the native API. There is no prompt injection or quality degradation. One documented exception: open-weight models used inside Claude Code (GLM, DeepSeek, Kimi, MiniMax, Qwen) are served through a format bridge — the provider only offers them on the OpenAI-style API, so the gateway converts the wire format both ways. The model’s output and your billing are untouched; only the envelope changes.

Question 10

How do I know which model actually served my request?

Accepted Answer

Check the response, not the model’s own words. A model can’t reliably state its own version — asking “are you Fable 5?” is unreliable for every provider, since a model released after its training cutoff won’t recognise its own name. The authoritative signals: every API response carries a model field naming the exact model that served the request, passed through byte-faithfully from the provider (we can’t alter it); your usage history shows the precise model billed on each request at its listed rate; and in the chat app the model selector labels every reply. Those are how you verify — not the chatbot’s self-description.

Question 11

Which coding agents are supported?

Accepted Answer

Claude Code, Codex, Cursor, Kilo Code, and Hermes all work today. Each connects by pointing its base URL at the gateway and using your InferenceHub key — see the Docs for copy-paste configs. Claude Code reaches both Claude and the open-weight daily drivers (GLM-5.2, DeepSeek, Kimi, MiniMax, Qwen); Cursor now reaches Claude too (sonnet, opus, haiku and more as custom models, Agent mode included); and Kilo Code reaches the full catalog — Claude, open-weight, and GPT-5.6 — from one config file.

Question 12

Can I use one key across multiple tools and models?

Accepted Answer

Yes. One unified API key works across every supported agent and every model. Switch models or tools without managing separate provider accounts or keys.

Question 13

Which API formats do you support?

Accepted Answer

Three native wire formats: the Anthropic Messages API (/v1/messages), the OpenAI Chat Completions API (/v1/chat/completions), and the OpenAI Responses API (/v1/responses). Use whichever your existing client already speaks.

Question 14

Do you train on my prompts or completions?

Accepted Answer

No. We do not use your prompts or completions to train models, we do not sell them, and we do not store them — content is processed in memory to route and meter your request, then transmitted only to the upstream provider that serves the model you selected, solely to fulfill your request. Our logs record request metadata (model, token counts, status), never content.

Question 15

What data do you store?

Accepted Answer

We store your account email and a hashed credential, your credit balance and transaction metadata, and per-request usage metadata (timestamp, model, token counts, status) needed to meter and bill usage. We do not store full card numbers, and we do not store the content of your prompts or completions. If you use the hosted chat app, your conversation history is stored so your sessions persist — you can delete conversations in the app at any time.

Question 16

How do I manage and revoke API keys?

Accepted Answer

API keys are scoped and can be created and revoked at any time from your account in the portal. If a key is ever exposed, revoke it and generate a new one — there’s no limit on rotation.

Question 17

Can I delete my account?

Accepted Answer

Yes. You can request deletion of your account at any time. We retain certain billing records as required for tax, accounting, and legal obligations — see our Privacy Policy for details.

Frequently asked questions

Pricing & billing

Models & integrations

Privacy & account

Still have questions?