DoroAI Models 2026: Every Supported LLM Tested — Codex budget users

The 30-second summary

In-depth review

DoroAI lists 2 models: GPT-4o and Claude 3.5 Sonnet. This is a bare-bones relay aimed at developers who just want Codex access without paying for a full-featured platform.

Model Breakdown

GPT-4o

This is the standard GPT-4o variant — not the turbo or mini release. Context window hits 100,000 tokens, which matches OpenAI’s official spec. In practice, that means you can dump a full codebase into a single prompt without hitting truncation.

Speed is the tradeoff. With only 94% uptime and a composite score of 56.4/100, you’re not getting the 99.9% reliability of direct OpenAI API access. Expect occasional timeouts during peak hours. For side projects or personal coding assistants, it’s fine. For production pipelines, you’ll want a backup relay.

Claude 3.5 Sonnet

Anthropic’s mid-tier model is the better choice here if you’re doing complex reasoning or multi-step code generation. It handles 100K tokens natively, and the safety rating of 3/5 suggests basic content filtering — less aggressive than Claude’s own API, but still present.

The catch: Claude 3.5 Sonnet through DoroAI routes through a third-party relay, so latency is higher than direct Anthropic access. Expect 3-5 seconds for first token on complex prompts.

Pricing

No promo code available. Minimum recharge amount is not published, which is unusual for Chinese relay stations. Expect at least ¥10-¥20 if they follow industry norms.

Pros & Cons

Pros

• Free trial exists — test before committing
• 支付宝 and 微信支付 accepted, no VPN needed
• 100K context window on both models
• Codex budget pricing (cheaper than direct API for heavy users)

Cons

• Only 2 models — no Gemini, DeepSeek, or local Chinese models
• 94% uptime is below average for relay stations (most aim for 97%+)
• No refund policy stated
• Minimal community feedback — hard to verify reliability claims
• Composite score of 56.4/100 signals quality concerns

Verdict

DoroAI is for one specific use case: you need GPT-4o or Claude 3.5 Sonnet for Codex-style coding, you want to pay with 支付宝/微信支付, and you don’t care about model variety or uptime guarantees.

Skip it if you need Gemini, DeepSeek, or any model outside these two. Skip it if your work depends on consistent uptime. The 94% uptime and 56.4 composite score are red flags for anything beyond hobby projects.

The free trial is the smart play. Register, run your standard code generation benchmarks, and see if the speed and reliability meet your threshold. If they do, the pricing is competitive. If they don’t, move on — there are better relays with broader model support.

FAQ

Q: Can I use DoroAI without a VPN in China? A: Yes. That’s the entire point. Access via 支付宝 or 微信支付, no VPN required for either payment or API calls.

Q: What’s the actual cost per token for GPT-4o? A: Not published. You’ll need to register and check the pricing dashboard. The free trial lets you test without paying.

Q: Is the 100K context window real or throttled? A: It’s real based on the platform data. Both models support 100K tokens. Whether the relay actually delivers full context without truncation depends on their backend — test with a large file during the free trial.

Pricing breakdown

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

Supported models

2 models across major vendors.

Frequently asked questions

Should you use DoroAI?

Codex budget users