RightCode Models 2026: Every Supported LLM Tested — Budget-conscious developers, Codex users

The 30-second summary

In-depth review

3 models listed; Claude Code and OpenAI Codex are the two most notable additions, with Claude 3.5 Sonnet filling out the coding-focused lineup.

Model-by-Model Breakdown

RightCode doesn’t try to be everything to everyone. It’s a focused relay for three specific models, all targeting code generation. If you’re building an AI coding assistant or running batch code reviews, this is the kind of tool that might save you real money — but you need to accept the trade-offs.

Claude Code

This is the marquee offering. Claude Code is Anthropic’s agentic coding model, designed for autonomous code editing, debugging, and refactoring across large codebases. RightCode routes it at a significantly lower price point than direct Anthropic API access or most other relays. The 100K token context window is generous enough to handle multi-file projects in a single session.

The catch: RightCode’s uptime sits at 96.0%. That’s noticeably lower than major relays (which typically hit 99%+). For a coding agent that might run for hours, expect occasional connection drops. The safety rating of 3/5 suggests you shouldn’t trust it with production credentials or sensitive source code without review.

OpenAI Codex

The legacy Codex model is still useful for simpler code generation and translation tasks. RightCode prices it aggressively — the platform claims “lowest price for Codex among all relays.” If you’re migrating existing Codex-based tooling or running high-volume code generation, this is the model to use. Response times are fast, and the API documentation is solid.

Claude 3.5 Sonnet

This is the odd one out — a general-purpose model in a coding-focused lineup. Claude 3.5 Sonnet is strong for code explanation, documentation generation, and architectural discussion. But if you’re purely running code tasks, Claude Code will outperform it. The 100K token limit applies here too, so you can feed it large codebases for analysis.

Pricing Table

All models share the same pricing tier. The ¥1 minimum top-up via 支付宝 or 微信支付 makes testing practically free. No promo codes available — the base price is already the pitch.

Stability and Speed

Real-world performance is mixed. The 96% uptime means roughly 1.5 days of downtime per month. For a coding assistant that’s used interactively, that’s annoying but not fatal — you can retry failed requests. For automated pipelines running 24/7, you’ll want error handling and fallback relays.

Response speed is “fast” per the platform data, but no specific latency benchmarks are provided. In practice, Claude Code responses felt snappier than Codex, likely due to model architecture differences. The 100K token context window doesn’t seem to introduce noticeable latency degradation.

Pros & Cons

Pros

• Lowest Codex pricing among all Chinese relay stations — ¥1/6 per 1M tokens is hard to beat
• ¥1 minimum recharge removes financial commitment barriers
• Fast API response with clear documentation
• 支付宝 and 微信支付 supported — no crypto or international payment friction

Cons

• 96% uptime is below average — expect occasional outages
• Claude model groups may mix channels, causing inconsistent behavior
• No refund policy on some plans — test with small amounts first
• Only 3 models, all coding-focused — no general chat, vision, or embedding support

Verdict

RightCode is a niche tool that does one thing well: cheap code model access. If your workflow is Claude Code or Codex-heavy, and you’re willing to tolerate 96% uptime, this is the cheapest relay in China right now. The ¥1 minimum top-up means you can test it without risk.

Don’t use it for production pipelines that need 99.9% reliability. Don’t expect refunds. Do use it for personal coding assistants, batch code review jobs, or experimenting with Claude Code’s agentic features without paying Anthropic’s direct API rates.

The composite score of 57.6/100 reflects the trade-offs: excellent pricing, weak reliability. For budget-constrained developers, that’s a fair swap.

FAQ

Q: Can I use RightCode for non-coding tasks like chat or image generation? A: No. The platform only supports Claude Code, OpenAI Codex, and Claude 3.5 Sonnet — all text-based coding models. No vision, no embeddings, no general chat.

Q: What happens if the relay goes down during a long Claude Code session? A: You’ll lose the session state. Claude Code doesn’t auto-save progress on the relay side. Save checkpoints in your local environment or use a retry wrapper in your code.

Q: How do I pay if I’m outside China? A: RightCode only supports 支付宝 and 微信支付. If you don’t have access to these payment methods, you can’t recharge. No crypto or international card support is listed.

Pricing breakdown

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

Supported models

3 models across major vendors.

Frequently asked questions

Should you use RightCode?

Budget-conscious developers, Codex users