Answers without the guesswork.

Create an account, activate access with a subscription, pay-as-you-go balance, or valid community voucher, then generate a key from API Keys. New accounts do not receive an automatic token allowance.

Use https://api.g0i.ai/v1 for OpenAI-compatible clients. Anthropic clients use https://api.g0i.ai with the /v1/messages path. Always follow the exact setup guide for your tool.

Open Models & Pricing or call GET /v1/models. Copy the exact ID; capitalization, punctuation, slashes, and hyphens matter.

Usually yes. OpenAI-compatible SDKs work with the /v1 base URL, Anthropic clients can use /v1/messages, and Responses-compatible clients can use /v1/responses.

The public catalogue follows GET /v1/models. A model is removed when it is no longer published by the live gateway, even if historical pricing metadata still exists.

Yes on compatible models. Support can differ by model family, so test the exact published model ID before relying on tool calls, structured output, vision, or other advanced features.

Yes. Use /v1/images/generations for compatible image models or the generation job endpoints for supported image and video workflows.

Claude Code should use the Anthropic base URL https://api.g0i.ai and the native /v1/messages format. See the Claude Code integration guide for environment variables.

Codex-compatible setups use https://api.g0i.ai/v1 and the Responses API. See the Codex integration guide for the current configuration pattern.

The public options are Pro, Pro Plus, Pro Max, and Pay-As-You-Go. Plan names, prices, allowances, and active promotions are loaded live on the Pricing page.

Add a USD-denominated balance and pay the published rate for each model. Input, output, cache-read, and cache-write tokens can have different prices. The current minimum top-up is shown on the Pricing page.

First-time customers start with a pay-as-you-go top-up of at least $5 and cannot buy a plan until that first payment settles. Existing customers (anyone with a completed payment on record) top up from $50 or buy a plan. Admin credits, vouchers and welcome credits do not count as a payment.

Your first successful request starts a five-hour window. Usage accumulates during that window, then the full allowance resets together five hours later.

No. New accounts start without active access. Community vouchers may be shared separately, but they are not a guaranteed signup benefit.

When a model exposes cache billing, cache writes and cache reads appear as separate per-million-token rates in Models & Pricing. A dash means no separate public cache rate is currently listed.

Confirm the key starts with sk-, has not been revoked, contains no spaces, and is being sent as Authorization: Bearer sk-.... Also verify the key is configured in the same shell or application process making the request.

The account may have no active plan, an expired voucher or subscription, or insufficient pay-as-you-go balance. Check Billing while signed in.

Refresh GET /v1/models and copy an exact current ID. Do not rely on an old screenshot, cached list, display name, or historical model alias.

Retry with exponential backoff and keep streaming timeouts generous for long reasoning requests. If the issue repeats, include the time, endpoint, and request ID when contacting support—never include your secret key.

The published policy says prompt and response content is not stored as ordinary product data. Lightweight operational metadata may be retained for billing, abuse prevention, and reliability.

Yes. Open API Keys while signed in and revoke the affected key. Create a replacement immediately if the old key was exposed.

Eligibility depends on the product and usage state. Unused vouchers and early plan cancellations may qualify under the published Refund Policy; submit requests with the relevant invoice or voucher reference.

Contact support from the account email or verified community identity and request account deletion. Support may verify ownership before removing account-associated data.