TokenHub
Access DeepSeek/Qwen/GLM without Chinese ID
4 followers
Access DeepSeek/Qwen/GLM without Chinese ID
4 followers
TokenHub gives you access to DeepSeek, Qwen, GLM and 50+ Chinese AI models through one OpenAI-compatible API. No Chinese phone required, no KYC, pay-as-you-go. Why: Official Chinese AI APIs require a +86 phone number to register. Pricing: • DeepSeek V3: $0.27/M tokens (10x cheaper than GPT-4o) • GLM-4-Flash: $0.01/M tokens • MiniMax: $0.20/M tokens Setup takes 1 minute — register at t-hub.cc, get a token, change base_url. That's it. Your existing OpenAI SDK code works unchanged.
Here's how TokenHub works under the hood, since a few people asked:
Stack:
One API (Docker) — unified gateway across 6 upstream providers. 50+ models behind one base_url.
Nginx — HTTPS via Cloudflare, rate limiting, scanner blocking. Unauthenticated requests get 401 at Nginx before touching the backend.
Flask (Python) — order management, auto token delivery, webhook integration with LemonSqueezy for credit card payments.
Vultr ($12/mo) — dual Dallas + Miami servers, 30s health checks, 90s failover.
Upstream providers:
DeepSeek (V3, V4 Flash, V4 Pro), Zhipu (GLM-4 series), MiniMax, Alibaba Bailian (Qwen). OpenRouter coming soon for GPT-4o, Claude, Gemini.
What I'd do differently:
Start with OpenRouter from day one — direct provider registrations took weeks
Build the pricing calculator earlier — tweaking model ratios manually was painful
Add webhook auto-delivery before launch, not after
Started as a weekend project to solve my own problem. Happy to answer technical questions.
TokenHub just got a massive upgrade — GPT-5.5 and Claude Opus 4.8 are now live.**
When I launched TokenHub on PH two weeks ago, the biggest feedback was "where's GPT-4o and Claude?" Fair. Today that changes.
What's new:
- GPT-5.5, GPT-4o, GPT-5, GPT-5-nano — full OpenAI lineup
- Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5 — full Anthropic lineup
- Pricing is transparent: cost + 10% margin only. No hidden multipliers
- First 100 users: $0.10 for 1M tokens
What hasn't changed:
- Still no Chinese phone number required
- Still WeChat & Alipay accepted (alongside USDT & credit cards)
- Still 400+ models behind one API endpoint
- Still $12/month servers (I run this solo)
The stack:
One API gateway → Nginx → Flask payment system → LemonSqueeze/PayPal/WeChat/USDT. Models sourced from official APIs and verified daily. Monitoring dashboard, usage analytics, pricing calculator — all built from scratch.
Honest numbers:
- 283 currently active models
- 9 completely free models (Gemma, Kimi, GLM, Mistral, etc.)
- $0.1 entry plan (1M tokens)
What's next:
Building a TikTok channel showing the behind-the-scenes of running an AI API business solo. And adding more payment methods.
Happy to answer questions about the architecture, pricing, or why I started this. Ask me anything.
What's new in June:
400+ AI models now available (Llama 4, Mistral Large, Gemma 4, Nemotron 550B, DeepSeek, Qwen, GLM, Kimi + 390 more)
Credit card payments via LemonSqueezy + PayPal
New pricing from $0.26/M (Starter $0.69, Pro $26 for 70M tokens)
6 upstream providers with automatic failover
Try the free models: google/gemma-4-31b-it:free, nvidia/nemotron-3-ultra-550b-a55b:free, moonshotai/kimi-k2.6:free
One API key. OpenAI SDK compatible. No Chinese phone. https://t-hub.cc
We just dropped prices by 54%! DeepSeek permanently cut API costs by 75% and we're passing it through. Starter now $0.69, Pro $26 for 70M tokens ($0.37/M).
Major Update: TokenHub now supports 400+ AI models — Llama 4, Mistral Large, Gemma 4, Nemotron 550B, DeepSeek, Qwen, GLM, plus 390+ more. One API key. OpenAI SDK compatible. From $0.26 per million tokens. No Chinese phone required. https://t-hub.cc
TokenHub now does 400+ AI models behind one endpoint. We started as a weekend project after getting scammed buying API keys on Telegram. Three months later: Llama 4, Mistral, DeepSeek, Qwen, GLM — all OpenAI compatible. Free models available (Gemma 4, Nemotron 550B, Kimi). From $0.26/M. Credit cards, PayPal, USDT accepted. https://t-hub.cc