🎁 618 Invite friends — they get 10% off every order, you earn 10% cashback Invite →

One Subscription. Unlimited Tokens, Every Major Model.

One key for the full Claude / GPT / Gemini / Grok chat lineup — works with Claude Code, Codex CLI and any SDK, plus exclusive Remote Session

TRIAL
¥1 one-time

Trial — taste everything first

40 real requests · all models · no weekly cap · use-and-done, no renewal

Register & Activate
NEW

Pro

For getting started

59 CNY/mo

≈ $8 USD/mo

  • Unlimited tokens
  • 500 / 5h · 2,500 / week
  • 2 API keys
Get Started
Recommended

Max

For daily heavy use

249 CNY/mo

≈ $35 USD/mo

  • Unlimited tokens
  • 2,000 / 5h · 10,000 / week
  • 10 API keys
  • Priority support
Get Started
Enterprise

Ultra

For professionals & teams

979 CNY/mo

≈ $137 USD/mo

  • Unlimited tokens
  • 6,000 / 5h · 40,000 / week
  • 20 API keys
  • All advanced features · per-key stats
Get Started

Models included in every plan

Any plan tier can call all models below. Every request draws from the same plan quota (1 request = 1 count) — no extra per-token charges.

Claude series — full lineup

The primary lineup for Claude Code — latest flagship included

claude-fable-5 NEW claude-opus-4-8 FLAGSHIP claude-opus-4-7 claude-sonnet-5 claude-sonnet-4-6 claude-haiku-4-5

Historical Opus / Sonnet / Haiku versions are also accepted. Model IDs go straight into the model field of Claude Code or any SDK.

GPT series

gpt-5.6-sol gpt-5.6-terra gpt-5.6-luna gpt-5.5 gpt-5.4 gpt-5 gpt-4o o3 o1

Gemini

gemini-3.1-pro gemini-3.1-flash gemini-3.5-flash

Text chat models

Grok

grok-4 grok-4-fast grok-3

Text chat models

The authoritative list of subscription-available models is returned by GET /v1/models on the gateway.

Configure Claude Code in One Click

Works with all Claude Code terminals and IDE plugins

irm llmapi.pro/setup.ps1 | iex

FAQ

Which models can I use with a subscription? Is everything in the model catalog included?

A subscription covers the full list shown in "Models included in every plan" above: the entire Claude chat lineup (including claude-fable-5) plus GPT / Gemini / Grok chat models, all sharing your plan quota. Models with -aws / -5x or -codex suffixes, GLM / DeepSeek direct-connect, and image/multimodal models are metered-API only and cannot be called with a subscription key — see the live catalogue on the API page (/channels). The authoritative list is returned by GET /v1/models.

How is this different from official Claude?

Your Claude Code experience is identical — same CLI, same tools, same workflow. We optimize backend infrastructure to dramatically reduce cost, and add exclusive features like Remote Session.

What happens if I upgrade from Pro to Max?

Upgrades take effect immediately to Max. Your remaining Pro time is Upgrades take effect immediately to Max. Your remaining Pro time is . After Max expires, you automatically revert to your remaining Pro days; if Pro also ends, you drop to Free. Example: You upgrade to Max for 1 month with 20 days of Pro remaining — you use 30 days of Max first, then 20 days of Pro.

Any catch with "unlimited tokens"?

We don't charge for or separately cap total token volume; instead we differentiate plans by a usage quota (5-hour and weekly rolling windows), weighted by the tokens each request processes (~1 per 10K tokens) — the same model as Claude Code's official subscription. Normal coding usage rarely hits any limit.

How is a "request" counted? How long does it last?

Quota is metered by usage, not a flat count of messages. A "request" is one API call from Claude Code to us, but how much it costs depends on the tokens processed that turn: roughly 1 per 10K tokens, a minimum of 1 for tiny requests, cached repeated context counted at 1/10, capped at 50 per request. Failed requests are never charged — timeouts, retries, and upstream errors (e.g. rate limits / 5xx) consume no quota; only requests that successfully return content count. This works the same way as Claude Code's official subscription — metered by token usage, with 5-hour and weekly windows. So a Claude Code request carrying a large 100K+ token context typically counts as 5–6, while a short Q&A counts as just 1.

Typical consumption patterns (for reference only; varies significantly based on task complexity):

  • Simple Q&A / single edit: ~ 1
  • Medium task (3–5 tool calls): ~ 2–4
  • Large-context Claude Code (deep refactor / long session): ~per turn5–6 higher for big changes

Based on the 5-hour window:

  • Pro (500 / 5h): Plenty for light-to-moderate daily use; ~1–1.5 hours of heavy Claude Code
  • Max (2,000 / 5h): Comfortable for a full day of moderate coding; ~half a workday of heavy Claude Code
  • Ultra (6,000 / 5h): Large teams / sustained multi-agent workloads

Quotas reset automatically on two independent rolling windows: 5-hour and 7-day. No overage charges, and failed requests are never counted — just wait for the next window. Claude Code will show a "rate limit" warning near the cap, then you're good to go again.

How to configure Claude Code?

Two steps:
export ANTHROPIC_BASE_URL=https://api.llmapi.pro:99
export ANTHROPIC_API_KEY=your_key
Then run normally claude That's it.