Poixe AI Review: Pricing & Comparison
Operating since 2024, 17+ providers / 8 model families covered, free models + level-based discounts
Last verified: 2026-08-11 · Visit official site →
Operating since 2024: what that means in the relay market
AI API relays are a high-attrition business. Between 2023 and 2024, as ChatGPT and Claude went mainstream, a wave of relay platforms sprang up — but a substantial share of them shut down within six months to a year, done in by funding problems, technical issues, or upstream API restrictions.
Poixe AI has been operating continuously since 2024, and that alone is a positive signal. It suggests:
- The platform has survived the early high-attrition period
- It has a stable user base and cash flow supporting operations
- The team remains committed to sustaining the business
That’s not a sufficient condition for choosing a platform on its own, but it is one necessary condition.
The level-based discount system: higher level, lower price
Poixe AI uses a user-level discount mechanism: new users start at Level 0 and auto-upgrade with cumulative top-ups and usage. The higher your level, the cheaper the same model becomes, and the higher your rate limits and concurrency. Billing is: per-request cost = standard price × level discount. All top-ups are settled in USD.
For enterprise or team users making frequent calls, this mechanism effectively encourages “top up and use more to earn discounts.” Levels auto-upgrade with cumulative top-ups — no application needed; higher level means more discount.
Free models and the traffic-owner referral plan
Poixe AI offers free model API calls to all users: append the :free suffix to a model name (e.g. gpt-4.1:free). Free models come in two groups — “big cup” (high-performance) and “small cup” (lightweight) — with independent quotas refreshed daily. They carry frequency, count, and context-length limits (e.g. Level 0 users get about 2 RPM / 5 RPD on the big-cup group), so they’re best for testing and light use.
The platform also runs a traffic-owner referral program: refer friends via a dedicated invite link and earn 2.5%-5% commission on their consumption, with each successful signup earning a $2.5-$5 new-user gift; earnings are withdrawable once you reach $14, via Alipay or USDT (TRC-20).
17+ provider coverage: what it’s good for
Poixe AI’s coverage of 17+ model providers across 8 model families (ChatGPT, DeepSeek, Claude, Gemini, Grok, Doubao, Qwen, Kimi) — while supporting three protocol families, OpenAI, Anthropic, and Gemini (switch by changing base_url and api_key) — concretely means that, under a single account, you can call:
- Text generation: Claude (all series), GPT, Gemini, Grok, DeepSeek, Doubao, Qwen, Kimi, and more
- Provider examples: OpenAI/Azure, Anthropic, Google, xAI, DeepSeek, Moonshot, SiliconFlow, Together, Fireworks, etc. (check the official site for the exact list; currently focused on text-chat models)
For teams that need to run comparative evaluations across multiple models, or dynamically switch models based on task type, this depth of coverage handles most common scenarios.
How to connect
Poixe AI is OpenAI-protocol compatible:
from openai import OpenAI
client = OpenAI(
api_key="your Poixe API key",
base_url="https://api.poixe.com/v1" # check the official site for the current endpoint
)
response = client.chat.completions.create(
model="claude-opus-4-8",
messages=[{"role": "user", "content": "Hello"}]
)
Information verified 2026-08-11. Membership-discount tiers, free-model quotas, and model coverage should be confirmed against Poixe AI’s official site in real time.
Related reviews
- API Yi — an old-guard aggregator, 300k monthly visits, multiple frontier models: an aggregation platform with an even longer operating history
- LinkAi — big bonuses on large top-ups: another option with a similar membership-discount mechanism
- KoalaAPI — stable relay, usage analytics: a relay platform with usage-management features
Related reviews
- Alibaba Cloud Bailian: Alibaba Cloud’s official AI platform, enterprise-grade SLA, the self-developed Qwen model series
- OpenRouter: a cross-vendor multi-model aggregation platform, 200-300+ models, one API key to call OpenAI/Anthropic/Google/Meta/Mistral and more
- Eden AI: a single API aggregating OpenAI/Anthropic/Google, with built-in fallback routing
- Meshs One: an international node for accessing domestic large models from overseas, AI API Gateway architecture
Quick facts
| Pricing model | Multi-protocol pay-as-you-go (USD); higher user level = bigger discount; free models (:free) with daily quotas; traffic-owner referral plan; check the official site for exact pricing |
|---|---|
| Model coverage | 17+ providers / 8 model families aggregated (ChatGPT/DeepSeek/Claude/Gemini/Grok/Doubao/Qwen/Kimi) across OpenAI, Anthropic, and Gemini protocols |
| Latency / SLA | Mainland direct connect; actual latency depends on real-world testing |
| Mainland direct connect | Direct connect |
| Best for | Developers / Enterprise |
| Referral program | A 'traffic-owner' referral program is live: 2.5%-5% commission on referred consumption, $2.5-$5 signup gift per referred new user, withdrawable once earnings reach $14 (Alipay/USDT). |
Pros
- Operating since 2024 — a track record that makes it more reliable than brand-new sites
- 17+ providers / 8 model families covered via a unified multi-protocol API, so one platform meets a wide range of AI needs
- Level-based discounts plus free (:free) models with daily quotas — high-frequency users see costs decline and new users test at zero cost
Cons
- Relatively niche, with limited independent review coverage
- Free models carry frequency/count/context-length limits, suited to testing and light use
- Little public latency-test data is available
Compare more AI API relays
See the full comparison board — filter by price tier, model coverage, and mainland direct-connect status.
Back to the comparison board →