Private beta

All the top AI models. No hidden limits.

Chat and API access to Claude, GPT, Grok and DeepSeek — one subscription, numeric limits, no hidden fair use.

Powered by Anthropic OpenAI xAI + Gemini, DeepSeek soon
 
This answer 0 credits

Switch models mid-conversation — pricing follows automatically.

Every top model, one place

Top models from Anthropic, OpenAI, xAI and DeepSeek — Fast, Balanced and Power tiers — behind one subscription, one chat, one API key.

Limits in numbers, not promises

Every cap is a number you can see: credits per model, a 5-hour window, requests per minute. No hidden "fair use."

One price, no surprises

Starter at $9.99 or Pro at $19.99 a month — fixed price, fixed credits, no metered bill at the end of the month.

Built not to fail mid-answer

If a model is briefly unavailable, we switch to an equal-or-cheaper alternative before the first token is sent — never mid-response.

Models

One catalog. Every top model. One API key.

Fast models for quick, cheap tasks. Balanced models for everyday coding and writing. Power models for the hardest problems. Switch between them in the same chat or the same API call — no separate accounts, no separate bills.

Fast
  • Claude Haiku 4.5 ≈171 credits / message

    Quick answers and everyday tasks, at the lowest cost in the catalog.

  • GPT-5.4 Nano ≈39 credits / message

    Fast, lightweight tasks where speed matters more than depth.

  • DeepSeek V4 Flash ≈16 credits / message

    The most economical model in the catalog — simple tasks, high volume.

  • Gemini 2.5 Flash ≈71 credits / message

    Google's fast workhorse — quick answers with a huge 1M context.

  • Gemini 2.5 Flash-Lite ≈16 credits / message

    The most affordable Gemini — instant replies for everyday questions.

Recommended
Balanced
  • Claude Sonnet 5 ≈513 credits / message

    Coding, writing and reasoning — the balanced default for daily work.

  • GPT-5.4 ≈475 credits / message

    A general-purpose alternative to Claude — same tasks, a different provider.

  • Grok 4.3 ≈143 credits / message

    A different model family from xAI — useful for a second opinion.

Power
  • Claude Opus 4.8 ≈855 credits / message

    Complex, multi-step problems that need deeper reasoning.

  • Claude Fable 5 Pro ≈1,710 credits / message

    The top of the catalog — long-form writing and the hardest problems. Pro plan only.

  • GPT-5.5 ≈950 credits / message

    Harder reasoning and analysis from OpenAI's flagship line.

  • GPT-5.5 Pro Pro ≈5,700 credits / message

    OpenAI's most expensive model in the catalog, for tasks that justify the cost. Pro plan only.

Credit estimates assume a ~2,000-token prompt and ~500-token reply. Your actual cost is metered from the provider's usage and always shown per answer.

Model names are trademarks of their respective providers (Anthropic, OpenAI, xAI, DeepSeek) and are listed here to indicate catalog/API compatibility only. AI Hub is not affiliated with or endorsed by these companies.

Pricing

One price, no surprises

Credits convert to messages at each model's published rate: budget models get you hundreds of messages, flagship models tens. Exact cost is shown before you hit send.

Starter — $9.99/mo

$9.99 / month

120,000 credits / month ≈233 Claude Sonnet 5 messages

Request an invite

120,000 credits a month — roughly 700 messages on Haiku 4.5, or about 140 on Opus 4.8, mix as you like. Every model except Fable 5 and the *-pro tier. Window cap: 18,000 credits per rolling 5 hours, 60 requests/min, 2 concurrent streams.

18,000 cr / 5h window · 60 RPM · 2 streams

Recommended

Pro — $19.99/mo

$19.99 / month

240,000 credits / month ≈467 Claude Sonnet 5 messages

Request an invite

240,000 credits a month — double Starter. Full catalog, including Fable 5. Roughly 280 messages on Opus 4.8, or 140 on Fable 5. Window cap: 36,000 credits per rolling 5 hours, 300 requests/min, 4 concurrent streams.

36,000 cr / 5h window · 300 RPM · 4 streams

Top-up — from $10

Request an invite

10,000 credits per $1 — the same rate as your plan. Stays active for 12 months. Kicks in automatically once your monthly credits run out.

Prices shown before local taxes. Private beta is invite-only with a manual payment step — no card stored until public launch.

Transparency

No hidden fair use

Cursor's monthly limit burned out in three days of normal use. Perplexity and GitHub Copilot quietly cut limits and got called out for it. We publish the numbers instead.

  • Every model's credit rate is public — before you send a message, not after the bill.
  • Rate limits are exact numbers: credits per 5-hour window, requests per minute, concurrent streams — shown in your account, not buried in a policy page.
  • We don't sell "unlimited." We sell a fixed number of credits at a fixed price, so the math works out the same way for you and for us.
  • If a limit changes, we announce it before it happens — not after users notice.
18,000
credits per 5-hour window, shown live
0
hidden fair-use clauses
12
live models under one key
100%
usage visible in your dashboard

FAQ

What are credits?

Credits are the app's internal currency: 10,000 credits = $1 of value. Every model in the catalog has its own published rate — cheap models cost fewer credits per message, flagship models cost more. You always see the exact cost before you send a message.

How much does one message cost?

It depends on the model and length. As a rough guide (2,000 input + 500 output tokens): DeepSeek V4 Flash ≈16 credits, GPT-5.4 Nano ≈39, Haiku 4.5 ≈171, Grok 4.3 ≈143, Sonnet 5 ≈513, Opus 4.8 ≈855, GPT-5.5 ≈950, Fable 5 ≈1,710. The exact number for your message is shown before you send it.

What happens when my credits run out?

Requests stop with a clear message — no silent downgrade, no hidden throttling. Top up from $10 (10,000 credits = $1) and keep going immediately, or wait for your next monthly renewal.

Which models are available?

Models across three tiers — Fast, Balanced, Power — from Anthropic, OpenAI, xAI and DeepSeek, behind one API key and one chat. Fable 5 and the *-pro tier are Pro-plan only; everything else is available from Starter.

How do I use the API?

One OpenAI-compatible endpoint (`/v1/chat/completions`), authenticated with a Bearer API key you generate in your account. Point your existing OpenAI SDK or curl at our base URL and switch the `model` field — no new SDK to learn.

Is my data safe? Do you log my prompts?

No-logs by default: request and response bodies are not stored on our servers. Chat history in the web app is optional, encrypted (AES-256-GCM), and yours to delete at any time — deletion is immediate and permanent. We keep billing metadata only: which model, how many tokens, cost, status — never the content.

How do I cancel my subscription?

During the private beta, billing is handled manually — email {{SUPPORT_EMAIL}} and we'll cancel it, no further charges. Credits already on your account stay usable until they expire or your current period ends.

How does top-up work?

Buy credits separately from your subscription — from $10, at the same 10,000 credits = $1 rate as your plan. Top-up credits stay valid for 12 months and are used automatically once your monthly plan credits run out (plan credits are always spent first).

What's the difference between Starter and Pro?

Pro doubles the monthly credits (240,000 vs 120,000) and the 5-hour window cap (36,000 vs 18,000), raises the rate limit (300 vs 60 requests/min) and stream limit (4 vs 2), and unlocks Fable 5 and the *-pro tier, which Starter doesn't include.

How do I get support?

Email {{SUPPORT_EMAIL}} — during the private beta, every message is read and answered by the team directly, not a bot.

Ready when you are

Private beta is invite-only. Request access and we'll set you up with a plan and your first credits.

Request an invite