Look up what 852 AI models actually cost, then switch in one line

Yes. You call the same model ID (claude-opus-5, gpt-5.6-luna) through an OpenAI-compatible gateway. Same weights, same context window, same SDK. Only the billing endpoint differs.

It depends on the model, and we will not pretend otherwise. The Claude family and most of the GPT-5.6 family currently run at 50% of the documented official list price. GPT-5.6 Sol is 37.5% off input / 25% off output because OpenAI's own $4/$20 rate is promotional (valid at least through 2026-11-21). Some models run at or near list. Use the price index for the current number on the model you actually use, that page is synced from the gateway's pricing endpoint, and we flag every figure we could not verify.

Gateway pricing follows what the upstream vendors charge. When a vendor runs a promotion the gateway discount widens; when it ends, the price moves back up. This is why we publish dated, per-model prices instead of a blanket "X% off" banner.

No. Minimum top-up is $1 and there is no monthly minimum. This is the main difference from the prepaid credit bundles most competitors sell.

Card via Stripe, plus other methods depending on region. Balances are denominated in USD.

We publish the gateway price for them but do not claim a discount, because DeepSeek bills by peak and off-peak windows and cache hit/miss tiers (a single "X% off" number would be misleading). The value there is access: one account instead of registering with each Chinese vendor and sorting out local payment.

Yes. Both let you override the API base URL. See the setup guides.

You are routing through a third-party gateway, so you are trusting their uptime and their model routing rather than Anthropic's or OpenAI's directly. Read how it works before you move production traffic.