FAQ

Yes. Tell us roughly how many people will use it and when you want to start. That is enough to discuss editions and pricing.

By the number of users, the authorization period and the features you need; the quote follows. Deployment, training and support are not part of the product and are discussed separately.

The features are the same; deployment and minimum seats differ. Flagship runs in a Tencent Cloud shared VPC from 1 seat; Exclusive runs in a dedicated VPC with dedicated network access and enterprise plugins, from 100 seats.

Companies that need the service inside their own environment with strict data and network requirements. Private Enterprise has no usage limit and is priced by consultation.

The official documentation says Enterprise cannot be used after expiry. Review usage and seats a month ahead and allow time for internal approval.

Enterprise invoices are issued through the Tencent Cloud console. When buying through Easy AI, who issues the invoice and how follows the order agreement.

Individual developers can see available models and prices through the product entry on the model services page. Enterprise pricing depends on the model, monthly volume and settlement method, so contact us directly.

OpenAI Chat (/v1/chat/completions) and OpenAI Responses (/v1/responses); Claude, Gemini, DeepSeek and others also accept the Anthropic format (/v1/messages). The catalog marks which formats each model supports.

Usually only the base URL, key and model name. Before going live, test each feature you use, such as streaming, tool calls and structured output.

Take average input and output tokens from the usage fields in responses and multiply by monthly requests and prices. The guide on estimating tokens and monthly cost has a worked example; prices come from the product page or a formal quote.

An integration test can be discussed; scope, cost and call rate are agreed in advance. Use separate test credentials and permitted samples, and never pass production keys in chats or documents.

A credential-free resource description: model IDs, interface, capacity and limits, supply period, supply basis and settlement. Do not send accounts or keys in the first conversation.

No. Part of the capacity runs Easy AI’s own business; what is available to partners depends on actual resources and your workload. Owning a data center also does not mean third-party models run in it.

Workload type (inference or training), model size, compute and storage needs, network requirements and duration. Location, equipment, operations and SLA terms are agreed separately.

Yes, along with the workload, quantity, duration and acceptable alternatives. Availability has to be confirmed.

Estimate weights as parameters × bytes per parameter, then add KV cache and headroom; the guides include a worked example. The final configuration comes from a real load test.

Submit them together if you like, but describe them separately. They differ greatly in memory, storage and duration.

Agree data export, deletion scope and account revocation before starting, and keep a deletion record. Arrangements are confirmed item by item.