FAQ

We offer a supported model catalog with KRW billing and a PII masking option. We don't mark up token prices (0%), and a platform fee of 3% applies when you top up credits — cheaper than OpenRouter's payment fee (~5.5%).

For documented compatible endpoints, set api_key and base_url to use the OpenAI SDK. Check each model’s documentation for supported parameters and response features.

We start with the latest models from OpenAI (GPT), Anthropic (Claude), and Google (Gemini), and progressively add xAI (Grok), DeepSeek, Qwen, GLM (z.ai), Mistral, and Korean models (Solar, HyperCLOVA X). Full list →

The inference location depends on the model — global models (GPT, Claude, Gemini) are processed in the US, Qwen and GLM (z.ai) in Singapore, and DeepSeek in China. Note that even a model from a Korean company may run on overseas inference infrastructure (e.g., Upstage Solar is processed on US infrastructure). The destination country for each model's cross-border transfer is published in a table in Article 5 of the Privacy Policy. In addition, PleumRouter's service operations infrastructure (servers and databases) is currently located overseas (Singapore), so regardless of which model you use, your account and usage records are stored on overseas infrastructure. You can mask sensitive information in advance with per-API-key 'PII masking'.

You prepay credits (minimum ₩5,000). Card payment (single installment) is supported.

No. We only place an upfront hold on the estimated cost at request time, and if the call fails the hold is released in full. Only successful calls are settled based on actual tokens used.

If the call still fails after automatic retries (twice), a 502 is returned and you are not charged.

Deactivate it immediately from the dashboard. Setting a monthly budget limit and an allowed-model restriction per key in advance helps limit the damage.

You can create an Organization, invite members, and use organization-level credits and usage aggregation.