NekoCode Review 2026: Pricing, Models, and API Notes

2.5 months to get new user credit, and the MAX group actually lives up to its name on consistency.

The Pricing Reality

NekoCode doesn’t hide behind a free tier. You pay upfront, and the minimum top-up is ¥20 — not much, but there’s no trial credit to test the waters. The new user deal is ¥15 for ¥30 credit, which effectively halves your entry cost. That’s a solid deal if you’re ready to commit. Payment is frictionless for Chinese users: Alipay and WeChat Pay both work. No crypto, no international card — just scan and go.

Models & Capabilities

Only two models are listed: GPT-4o and Claude 3.5 Sonnet. That’s it. If you need Gemini, DeepSeek, or any other provider, this isn’t your relay. The max token context is 100,000 — sufficient for most code reviews, document analysis, or long conversation threads. It’s not the 200K+ some newer relays offer, but it’s enough for practical work.

China Access & API Compatibility

This is a relay station designed for Chinese developers. No VPN required. The API endpoint works with standard OpenAI-compatible clients (e.g., OpenCat, NextChat, or custom scripts). If you’ve used any other relay, the integration is identical — just swap the base URL and API key. The station is newer (2026 launch), so the infrastructure is modern but the community trust is still building. The 98% uptime is decent, not stellar. Expect occasional short outages during peak hours, but I didn’t see prolonged downtime in my testing.

Channel Quality: MAX vs Kiro

NekoCode runs two channel groups: MAX and Kiro. The MAX group is the headline — stable output, consistent response patterns, no random truncation or weird hallucinations. The Kiro group is cheaper but less reliable; I’d stick with MAX for production work. The older Claude top-ups (separate credit for Claude-only usage) are no longer available. Everything now runs through the unified balance. If you were relying on that, you’ll need to adjust your budgeting.

Safety & Output Reliability

Safety rating is 4/5. Outputs are filtered but not aggressively censored. For technical use (code generation, debugging, documentation), this is fine. If you need uncensored responses for sensitive research, look elsewhere. The 78.4 composite score reflects a capable but limited platform. It’s not trying to be everything — it’s focused on quality for a small model set.

Pros & Cons

Pros

• MAX group delivers genuinely stable outputs with minimal variance
• Clean interface, well-maintained — no cruft or broken features
• New user deal effectively gives ¥15 free credit
• Both MAX and Kiro groups available for price/quality tradeoff Cons
• Older Claude top-ups discontinued — unified balance only
• Minimum top-up ¥20 blocks casual testers
• Newer station (2026) — less community history to judge reliability

Verdict

NekoCode is for developers who value channel quality over model variety. If you mainly use GPT-4o and Claude 3.5 Sonnet, and you want consistent outputs without managing multiple subscriptions, this relay delivers. The ¥15 new user deal makes the entry cost negligible. Skip it if you need a wide model selection, a free trial, or if you’re price-sensitive about top-up minimums. For its niche — reliable MAX channel access for Chinese developers — it’s a solid choice.

FAQ

Q: Can I use NekoCode with NextChat or OpenCat? A: Yes. The API is OpenAI-compatible. Just set the base URL to NekoCode’s endpoint and use your API key. Works with most standard clients. Q: What happens if I hit the 100,000 token limit? A: Your request will be truncated. There’s no automatic fallback to a higher context model. Plan your prompts accordingly if you’re processing long documents. Q: Is the ¥15 new user deal permanent? A: It’s available at launch (2026). No expiration date has been announced, but these deals often change. Claim it early if you’re interested.

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.