FAQ
What is Nexotao?
A reseller for AI model access. With one account and one Rupiah balance, you can use top models for text — via web chat, API, or Claude Code, with no foreign credit card.
Is it safe?
Your requests are forwarded to the official model providers for processing. We do not sell your data. Details are in the privacy policy at nexotao.com.
Can my balance expire?
No. The balance does not expire unless stated otherwise, and remains usable until it runs out.
How do I pay?
Top up via QRIS. No subscription — you only pay for the balance you add, and it is deducted based on usage.
Can I use it for Claude Code?
Yes. Nexotao is compatible with the Anthropic endpoint, so Claude Code can be pointed at Nexotao and paid for in Rupiah. See Claude Code.
Are there refunds?
Top-ups are final and the balance is generally non-refundable. For cases such as double charges or system failures, see the refund policy at nexotao.com.
How much does it cost?
Live rates are on Models & Pricing. They are currently well below official pricing — treat it as a promo price.
Why does DeepSeek sometimes return 429 / “server at capacity”?
Because the request isn’t a DeepSeek job. DeepSeek-V4-Flash is positioned for
high-volume, cheap work — classification, extraction, short chat — and has a
small capacity of about 20k tokens/minute (TPM) per backend. A single
agentic/coding or large-context request easily exceeds ~50k input tokens,
overrunning one backend’s entire per-minute budget, so it returns
429 server at capacity, retry shortly — no number of parallel backends helps,
because a single request cannot be split across backends.
The fix isn’t to keep retrying, it’s to use the right model: for coding/agentic
and large context, select a Claude (Opus/Sonnet) model id directly (for example
claude-opus-4-8). DeepSeek stays the best pick for volume/cheap workloads.