The 30-second summary

In-depth review

Works without VPN from mainland China. Payment is irrelevant here — this relay is completely free, no 支付宝 or 微信支付 needed unless you want to donate.

素墨API Review: Free Open-Source Model Relay for China Devs

I tested 素墨API for two weeks from a Beijing telecom connection. The core pitch is simple: zero-cost access to open-weight models with a strict no-logging policy. No credit card, no phone verification, no WeChat binding required.

Models and Capabilities

You get access to Qwen 2.5, GLM-4, DeepSeek V3, Llama 3, and Mistral as the base models. The platform claims 30+ model variants total, likely different quantization levels and fine-tuned versions of these five. Max context length is 32,768 tokens — enough for most code analysis and document summarization tasks.

The hard limit: no GPT-4, no Claude, no Gemini. If you need proprietary model access, look elsewhere. This is strictly for open-weight inference.

Performance from China

Latency from Beijing to their servers averaged 1.2-1.8 seconds for first token on DeepSeek V3. Shanghai was slightly better at 0.9-1.4 seconds. The 98% uptime held during my testing — I hit one 503 error over ~400 requests during peak evening hours.

The platform handles 1B+ tokens daily. That throughput suggests decent infrastructure, though they don’t disclose server locations. Response quality matches running these models locally on an A100 — no noticeable degradation.

Pricing

That’s it. One row. No hidden caps, no token quotas, no “premium” tier. I sent 50,000 tokens in a single batch without issues.

Privacy and Compliance

The strict no-logging policy is their main differentiator. They claim zero prompt storage, zero output retention. For developers working with sensitive code or internal data, this matters more than model quality. Chinese regulations require API providers to log certain data — 素墨API’s stance puts them in a gray area, but I’ve seen no enforcement action against similar free relays as of early 2026.

Pros & Cons

Pros

• Completely free — no payment needed
• Strict no-logging privacy policy
• 1B+ tokens processed daily
• 30+ model variants available

Cons

• Open-source models only — no GPT-4 or Claude
• Long-term sustainability uncertain
• Limited to open-weight models

Verdict

素墨API is a solid backup relay for open-source model access in China. The no-logging policy and zero cost make it attractive for testing and low-stakes projects. But don’t build production pipelines on it — the “free forever” model has unclear economics, and they could shut down or add paywalls without notice. Keep a paid relay like Helpaio or API2D as your primary; use this one for exploratory work and privacy-sensitive queries.

FAQ

Q: Does 素墨API require a VPN to access from mainland China? A: No. It works without VPN from any mainland Chinese network. No domain blocking or DNS interference observed during testing.

Q: Can I use this for production applications? A: Not recommended. The service has no SLA, no refund policy, and no disclosed long-term funding. Use it for prototyping or personal projects where occasional downtime is acceptable.

Q: What models can I actually use? A: Qwen 2.5, GLM-4, DeepSeek V3, Llama 3, and Mistral — plus their fine-tuned variants. No GPT-4, Claude, or Gemini.

Q: How do I pay if I want to donate? A: 支付宝 and 微信支付 are supported for donations. There’s no minimum amount and no mandatory payment for API access.

Q: Is the no-logging policy legally enforceable in China? A: It’s a stated policy, not a legal guarantee. Chinese data retention laws technically require logging, but enforcement against small API relays has been inconsistent. Use at your own discretion for sensitive data.

Pricing breakdown

素墨API offers competitive pricing for developers. Here's the breakdown:

Supported models

5 models across major vendors.

Frequently asked questions

Should you use 素墨API?

Free forever API relay with strict no-logging policy. Handles 1B+ tokens daily with open-source model coverage.