RightCode lists exactly 3 models: Claude Code and OpenAI Codex are the standout additions, with Claude 3.5 Sonnet rounding out the set. This is a lean, coding-only lineup.
Model-by-Model Breakdown
OpenAI Codex
Codex is the primary draw here. RightCode claims the lowest price for Codex among all relay stations. At ¥1 (in) / ¥6 (out) per 1M tokens, it undercuts typical per-token costs by a wide margin. The 100,000 max token limit applies across all models, so you get the full Codex context window. Stability: Uptime sits at 99.3%. Not the highest in the market, but for a budget relay, it’s acceptable. The 548ms average latency is fast enough for interactive coding sessions. You won’t notice the delay. What’s missing: RightCode does not expose per-model SLAs or indicate which models degrade first during incidents. If Codex goes down, you’ll see it in the overall 99.3% number, but you won’t get a model-specific alert.
Claude Code
Claude Code is the second coding specialist. It shares the same token limit and API endpoint structure as Codex. Speed rating is 4.5/5, which aligns with the 548ms latency benchmark. The relay doesn’t throttle Claude Code differently from Codex — both run on the same infrastructure. Caveat: The platform data notes that “Claude groups may occasionally mix channels.” This means your Claude Code requests might route through a different backend channel than expected. For production pipelines, this introduces nondeterministic behavior. For personal use, it’s rarely noticeable.
Claude 3.5 Sonnet
Sonnet is the general-purpose fallback. If your coding task requires reasoning beyond what Codex or Claude Code handle well, Sonnet is available. It’s the same price tier as the other two models — no premium markup for Sonnet. Context window: All three models share the 100,000 token cap. RightCode does not offer tiered context windows. You get the same limit regardless of which model you pick.
Pricing
| Model | Input Price (¥/1M tokens) | Output Price (¥/1M tokens) | Context Window | Rate Limit |
|---|---|---|---|---|
| OpenAI Codex | ¥1 | ¥6 | 100K | Unlimited |
| Claude Code | ¥1 | ¥6 | 100K | Unlimited |
| Claude 3.5 Sonnet | ¥1 | ¥6 | 100K | Unlimited |
| No promo code is available. Payment is via Alipay or WeChat Pay. Minimum top-up is ¥1, which is the lowest floor I’ve seen — you can test the service for pocket change. | ||||
| Bottom line: If you only need Codex or Claude Code, RightCode is the cheapest option in 2026. If you need a broader model selection, look elsewhere. |
Pros & Cons
Pros
- Lowest Codex pricing among all relays — ¥1/¥6 per 1M tokens
- ¥1 minimum recharge — zero barrier to entry
- 4.5/5 speed rating; 548ms latency is snappy for code completion
- 100K context window across all models Cons
- Only 3 models — no GPT-4, Gemini, or DeepSeek support
- Claude channels may mix unpredictably
- No refund policy specified
- No per-model SLA or uptime breakdown
Verdict
RightCode is a specialist tool, not a general-purpose relay. If your workflow is 90% Codex or Claude Code and you want the lowest possible price, this is the pick. The 99.3% uptime and 548ms latency are good enough for daily use. The lack of model diversity and the channel-mixing issue on Claude are real downsides, but they don’t outweigh the cost advantage for budget-constrained developers. Skip RightCode if you need GPT-4, Gemini, or any non-coding model. Go with it if you’re fine with a three-model lineup and want to pay almost nothing per token.
FAQ
How does RightCode’s Codex pricing compare to other relays?
RightCode charges ¥1 in / ¥6 out per 1M tokens for Codex. As of 2026, this is the lowest listed price among all relay stations in this review batch. Most competitors charge 2-3x more for the same model.
Can I use RightCode for non-coding tasks?
Technically yes — Claude 3.5 Sonnet is a general-purpose model. But the platform is optimized for coding use cases. The documentation focuses on code completion, and the model selection is limited to code-oriented LLMs.
What happens if a Claude channel goes down?
RightCode does not provide per-model uptime data. The overall uptime is 99.3%. If a specific Claude channel fails, requests may route through alternative channels, but the platform data warns that channel mixing can occur. You won’t get a dedicated SLA for Claude Code vs. Sonnet.
Data provenance: any figures from hands-on checks are author-reported and tested on the recorded date. Independent verification is unavailable unless a source is linked.
