FAQ

The Models page shows reference prices before you integrate; actual charges follow your console records. Different providers, model versions, caching strategies and multimodal capabilities may map to different billing items.

Models that support caching are labeled in their model information. Cache read and write prices are recorded separately, and the console itemizes input, output, cache reads and cache writes wherever possible.

Standard API access is billed by actual usage. Enterprise plans can be configured with fixed budgets, payment terms or annual commitments — contact sales.

Multimodal models are usually billed by their own rules — per token, per image, per second of video, or per task. See the Models page and console records for specifics.

Requests that never reach model inference (auth failures, invalid parameters) usually incur no model cost. If the upstream finished inference but the transfer was interrupted, the corresponding consumption may apply — request status is shown in the console.

Streaming only changes how the response is delivered; billing still follows actual input/output usage.

Enterprise customers can request custom plans based on monthly usage, annual budgets and model mix.

The console will open online top-ups, bank transfer and enterprise payment terms over time.

Standard prepaid credits follow the rules shown in the console; enterprise contract quotas follow the contract.

Billing is denominated in USD (quota, 500000 = $1); CNY is an on-site display conversion (at the site settlement rate). Any on-site amount can be converted back to quota for verification, and the console's usage stats and request logs map one-to-one with charge records.

apiex keeps the necessary request metadata, usage and billing records. Retention policies for content are configured explicitly in the console and enterprise agreements.

Customer business data is not used for model training by default. Data used for troubleshooting or product improvement follows permission, anonymization and agreement terms.

The enterprise plan roadmap includes SSO, member management, key permissions, budgets and audit logs.

Most scenarios only need a Base URL, API Key and model ID swap. Complex toolchains can follow the CLI and SDK configuration examples in the docs.

Common OpenAI SDK calling conventions work out of the box, with configuration references for tools like Codex, Claude Code and Cursor.

Yes. Enterprise customers get integration assessment, model selection, usage accounting and troubleshooting support.