Terminal.Pub Cheapest Plan 2026: Price Breakdown & Cost Calculator — Pure budget testing, not production

The 30-second summary

In-depth review

GPT-4o input costs ¥0.5 per million tokens on Terminal.Pub, with output at ¥2.5 per million. That’s roughly 1/10th of what you’d pay going direct to OpenAI — and for Claude 3.5 Sonnet, the pricing is similarly aggressive. This isn’t a production-tier service; it’s a budget testing ground for developers who want to experiment with these models without burning cash.

Pricing Breakdown

Terminal.Pub runs on a pay-as-you-go model with no monthly subscription. You recharge via 支付宝 or 微信支付, and the platform explicitly labels itself as “pure budget testing.” The numbers back that up.

Compare this to OpenAI’s direct pricing: GPT-4o costs $2.50/MTok input and $10/MTok output. At current exchange rates (~¥7.2/USD), that’s ¥18 input and ¥72 output. Terminal.Pub is 36x cheaper on input and 28x cheaper on output. The gap is suspiciously wide.

There’s also a free API tier available — specific token limits aren’t disclosed, but it’s worth testing your workflow at zero cost before recharging.

Pros & Cons

Pros

• Pricing is genuinely absurd — ¥0.5/2.5 per million tokens for GPT-4o and Claude 3.5 Sonnet
• Free API groups let you evaluate before spending
• Supports 支付宝 and 微信支付, so no foreign credit card needed

Cons

• Platform launched February 2026 — barely a month old at time of writing
• 94% uptime is below what I’d accept for anything beyond hobby projects
• No subscription option; you’re forced to recharge and track usage manually
• Safety rating of 3/5 suggests you shouldn’t send sensitive data through this relay

Uptime and Reliability

94% uptime over the tracked period. That means roughly 1.5 days of downtime per month. For a testing playground, that’s tolerable. For anything customer-facing, it’s a non-starter. The platform is too new to have meaningful historical data — Helpaio scores are listed as N/A, which tells you nobody’s been watching this service long enough to benchmark it.

Hidden Fees and Batch Discounts

No batch discounts are advertised. The pricing is flat per-token regardless of volume. No hidden fees either — what you see per million tokens is what you pay. The refund policy is unspecified, so assume no refunds on recharged balances. Start with the free tier, test your workload, then recharge the minimum amount needed.

Verdict

Terminal.Pub is a risk-reward bet. The pricing is the lowest I’ve seen for GPT-4o and Claude 3.5 Sonnet through any Chinese relay station. But the platform is weeks old, uptime is mediocre, and the business model at these prices raises questions. Use it for prototyping, stress-testing prompts, or comparing model outputs. Do not route production traffic through it. If you need reliability, pay more elsewhere.

FAQ

Q: Is there a minimum recharge amount? A: The platform data doesn’t specify a minimum. Start with the free API groups first, then recharge whatever you’re comfortable risking.

Q: Can I use Terminal.Pub without a VPN from China? A: Yes — that’s the point of this relay. It routes API calls to OpenAI and Anthropic servers without requiring you to bypass the GFW directly.

Q: What happens if the platform shuts down with unused balance? A: The refund policy is not specified. Treat any recharged balance as non-refundable. Don’t load more than you’re willing to lose.

Pricing breakdown

Terminal.Pub offers competitive pricing for developers. Here's the breakdown:

Supported models

2 models across major vendors.

Frequently asked questions

Should you use Terminal.Pub?

Pure budget testing, not production