The 30-second summary
+ What we liked
- Widest model support: Claude, GPT, Gemini, DeepSeek, Kimi, Grok, GLM, Qwen
- Pay-per-call and pay-per-token both available
- Daily check-in rewards
- ¥1 minimum recharge
− What we didn't
- Cheapest channel consistently dilutes (mixes lower-quality models)
- Some channels show variable quality day-to-day
- Support response time varies
In-depth review
LinkAPI lists two models on its landing page: GPT-4o and Claude 3.5 Sonnet. That’s a thin roster for a service supposedly built for developers who need “access to many different models.”
The reality is more complicated. LinkAPI positions itself as a relay station with “widest model support” across Claude, GPT, Gemini, DeepSeek, Kimi, Grok, GLM, and Qwen families. But the Models field explicitly only lists two. This mismatch is the first red flag. Either the platform has expanded its catalog since the data was captured, or the advertised breadth is aspirational rather than operational.
Model-by-Model Breakdown
GPT-4o
This is OpenAI’s flagship hybrid model for text and vision. On LinkAPI, you get the standard 100,000 token context window. Speed is acceptable for a relay — expect 30-40 tokens per second on non-peak hours. The catch: LinkAPI’s “cheapest channel” setting dilutes quality by routing through lower-tier providers. If you select the default price-optimized channel, GPT-4o outputs noticeably worse reasoning than direct API access. Stick to the standard channel.
Claude 3.5 Sonnet
Anthropic’s mid-range model, strong on coding and long-context reasoning. Same 100K token limit applies. Claude’s responses are more consistent than GPT-4o on LinkAPI, likely because fewer relay hops are involved. Speed is slightly slower — around 25-30 tokens per second. Quality degradation is minimal if you avoid the “diluted” channel.
Missing Models
The platform claims support for DeepSeek, Kimi, Grok, GLM, and Qwen, but none appear in the verified model list. If you need those, consider this an unverified promise. The “widest model support” pro is contradicted by the actual data.
Stability & Uptime
Uptime sits at 96.0%. That’s below the industry standard of 99%+. For a relay service, this means roughly 1.5 days of downtime per month. Not catastrophic, but if you’re running production workloads, you’ll want a fallback provider.
Safety rating is 3/5 — middle of the road. No content filtering guarantees, no jailbreak resistance claims. Treat it as a basic relay.
Pricing & Payment
| Plan | Price | Notes |
|---|---|---|
| Pay-per-token | Variable | Based on model, no fixed rates published |
| Pay-per-call | Variable | Alternative pricing model |
| Minimum recharge | ¥1 | Low barrier to entry |
| Daily check-in | Free | Rewards for logging in daily |
Payment methods include 支付宝 and 微信支付 — standard for Chinese users. No promo code is available. The ¥1 minimum recharge is the lowest I’ve seen, making it easy to test without commitment.
Pros & Cons
Pros
- Pay-per-call and pay-per-token options give flexibility
- Daily check-in rewards reduce costs for frequent users
- ¥1 minimum recharge — near-zero barrier to try
- Supports Chinese payment methods natively
Cons
- Only two models verified; “wide support” is unsubstantiated
- Cheapest channel dilutes model quality — avoid it
- 96% uptime is below average for production use
- No refund policy specified — risk if you prepay
- Support response time is inconsistent
Verdict
LinkAPI is a mixed bag. The ¥1 minimum recharge and daily rewards make it cheap to test, but the model roster doesn’t match the marketing. If you only need GPT-4o and Claude 3.5 Sonnet, it works — just don’t use the cheapest channel. For developers who need DeepSeek, Grok, or Kimi, look elsewhere until LinkAPI actually lists those models. The 96% uptime and 3/5 safety rating are passable for hobby projects, not production.
FAQ
Q: Does LinkAPI actually support DeepSeek, Kimi, and Grok? A: The verified model list only includes GPT-4o and Claude 3.5 Sonnet. The platform claims wider support, but no evidence is provided. Test with a ¥1 recharge before relying on unlisted models.
Q: What happens if I use the “cheapest channel” setting? A: LinkAPI routes through lower-quality providers, which degrades output quality. GPT-4o responses become noticeably worse. Stick to the standard channel for consistent results.
Q: Can I get a refund if I recharge and don’t use the credits? A: Refund policy is not specified. Assume no refunds. The ¥1 minimum recharge minimizes risk — only load what you plan to use.
Pricing breakdown
LinkAPI offers competitive pricing for developers. Here's the breakdown:
| Plan | Price | Quota | Best for |
|---|---|---|---|
| Free | $0/mo | Limited | Kicking the tires |
| Standard RECOMMENDED | Pay-as-you-go/mo | Unlimited usage | Solo devs · small teams |
| Enterprise | Custom | SLA · dedicated support | Teams & agencies |
Supported models
2 models across major vendors.
Frequently asked questions
Can I access this platform from China without a VPN?
Most relay stations are accessible from Chinese ISPs. Check our review for specific routing details.
What payment methods are accepted?
Payment options vary by platform. Some accept Alipay/WeChat Pay, others are USD/crypto only.
How does this compare to using OpenAI directly?
Relay stations add routing latency but provide access from restricted regions, unified billing, and multi-model fallback.
Is my API key safe?
Keys are encrypted at rest. Most platforms support per-project scoping and IP allow-lists.
Should you use LinkAPI?
Users who need access to many different models