Originally published on my blog — read it there with the full context, code and updates.
Chinese frontier models — GLM-5.3, DeepSeek V4, Qwen3.8-Max, Kimi K3 — now
score close to Claude on coding benchmarks while costing a fraction of Western
frontier APIs. For a solo developer or a small business running AI automation,
that price difference is the difference between "AI that pays for itself" and
"AI that's too expensive to run."
But using them from Malaysia means crossing three walls first: payment rails,
passing mainland-China phone verification, and (on some platforms) real-name
(实名) registration. This is what I actually did, what broke, and what I'd do
differently.
Cost vs. Complexity: Why It's Worth Exploring
Chinese API pricing is often 3–10× cheaper than the equivalent Western model.
As of this writing, DeepSeek's V4.1 Flash is around US$0.15–0.30 per 1M
input tokens and US$0.60–1.20 per 1M output tokens (off-peak hours are half
price; cached input is ~20× cheaper still). A US$10 top-up goes a long way —
tens of millions of tokens for agent workloads if you route carefully.
The catch: most of these platforms were originally built for mainland developers. That
means Chinese payment rails and, frequently, a mainland phone number for SMS
verification.
Wall #1 — Payment Rails and Compliance
Payment settlement is the first barrier when accessing frontier models across borders. In practice, you must distinguish between personal sandbox experiments and enterprise production deployments:
- Personal Sandbox Testing: For individual developers validating prompts and workflows, international Alipay accounts allow linking foreign Visa/MasterCard credit cards for direct top-ups on supported gateways.
-
Enterprise Production Architecture: Production client systems must never depend on unofficial peer transfers or gray-market top-ups. Commercial systems demanding corporate invoicing, tax compliance, and auditable SLAs should adopt formal global routes:
- DeepSeek Global Platform: Accepts international credit cards in USD directly and generates corporate receipts without requiring mainland entities or phone numbers.
- BytePlus ModelArk: ByteDance's Singapore entity operates full enterprise USD billing and hosts inference data in Singapore/Malaysia regional clusters, satisfying corporate data residency requirements.
- Self-Hosted Gateway Failover (LiteLLM): Deploying LiteLLM on private infrastructure orchestrates multi-vendor routing, providing seamless failover across regional endpoints and fallback providers if a primary route degrades.
Wall #2 — phone verification
TokenRhythm (基元律动) — the gateway I use most — requires a mainland China
phone number for registration. There's no way around that requirement on their
side, and I won't recommend dodging it. If you don't have access to a mainland
number, platforms like DeepSeek (email signup) or BytePlus ModelArk (Singapore,
USD billing) let you skip this wall entirely — more on those below.
What made TokenRhythm worth the hassle:
-
One API key, OpenAI-compatible. It aggregates GLM-5.3, Qwen3.8-Max,
DeepSeek V4, Kimi K2.7-Code, MiniMax, and Seed models behind a standard
https://tokenrhythm.studio/v1endpoint. Any agent that speaks OpenAI-style APIs (OpenCode, Claude Code, LiteLLM, my own bots) just works. - Referral credits. Registering through an invite link gets you ¥68 (~US$10) in free credit; referring another registration gets you another ¥68. The reward unlocks after you install their desktop app (OpenSquilla) and make at least one call from it — after that, the same API key works in any agent.
- The credit is real money. ¥68 (about US$10) covered roughly a month of light agent work, or about a week of heavy sessions, with DeepSeek V4 Pro through the gateway. DeepSeek's pricing is genuinely very good, so the credit goes further than you'd expect.
The gotcha nobody warns you about
Promotional credit on these platforms is time-limited, not a balance.
TokenRhythm's new-user credit expires 30 days after it's granted — my first
batch expired on 2026-09-03 with some still unused. Don't register and hoard
the credit for "later". Spend it, or lose it.
Also worth knowing: DeepSeek is retiring its V4 Pro model on September 14,
2026 — after that, deepseek-v4-pro requests are served by V4.1 Flash and
billed at Flash prices. That's good news for cost, just don't expect the Pro
behaviour to stay the same (other replacements will come).
What actually works, in order of friction
- DeepSeek directly — email signup, no China phone, no real-name. Cheapest per-token of anything I use. The official docs list the current rates with a peak/off-peak split.
- TokenRhythm (基元律动) — best model menu for the money, OpenAI-compatible, but you need the mainland phone number and an invite link for the ¥68 credit.
- BytePlus ModelArk — ByteDance's international platform: Singapore entity, USD billing, no China phone, and data processed in Malaysia/Singapore region. Slightly pricier per token, far less friction.
- OpenCode Go — US$10/month flat for a curated roster of open coding models (GLM, DeepSeek, Kimi, Qwen, MiniMax) with roughly 6× the usage value. Not Chinese-billing at all, which is exactly its point.
What I'd do differently
- Check the verification requirements before signing up. Email-only platforms (DeepSeek) or international platforms (BytePlus, OpenCode Go) avoid the phone wall completely. Only reach for the mainland-phone platforms when their model menu or price genuinely beats the alternatives.
- Burn promotional credit immediately. It's a 30-day lease, not a balance.
- Start with a small recharge. Test the actual model on your real workload for a week before committing.
- Re-check prices before each top-up. Chinese API pricing moves fast — DeepSeek cut prices permanently this year, GLM runs periodic flash promos, and deals appear and vanish weekly. Never assume last month's price is this month's.
The result
I now run my AI automation — Telegram support bots, agentic coding sessions,
background LLM jobs — on Chinese LLM APIs for roughly the price of a coffee per
week of active use. The ¥68 TokenRhythm credit alone covered a month of light
use (a week of heavy work), and a US$10 DeepSeek top-up lasts even longer at
V4.1 Flash prices.
The walls are real, but they're payment-rail and verification walls — not
quality walls. If you're building AI features for a business that needs to
actually make money, the Chinese frontier APIs are worth the setup.
Want AI automation like this for your business?
I build AI-powered automations, Telegram/WhatsApp support bots, websites, and
self-hosted infrastructure for Malaysian businesses — and I set up the cheapest
LLM routing so your running costs stay small. If that could help your business,
I'd love to talk:
- 📱 WhatsApp: +60 12-797 2969
- 📧 Email: me@hoelee.com
- 🌐 Website: hoelee.com
I write about self-hosted infrastructure, DevOps and the failures behind them at blog.hoelee.com.
Top comments (0)