One API endpoint for Claude, OpenAI, Grok, GLM, Kimi, DeepSeek and Jev

TopxAI is a model API service. Models from seven lines are served through OpenAI- and Anthropic-compatible endpoints at fixed USD prices, on prepaid credit. TopxAI does not store prompts, completions or uploaded files.

Models

Model catalogue · 7 lines Claude Fable 5.2, Sonnet 5.5 and Opus 5.5; GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT Image 2.6 (Sunburst, Flare); Grok 4.7 and Grok Imagine Video 1.5; GLM-5.3-Abliterated and GLM-5.3-Flash-Abliterated; Kimi K3.1 Pro; deepseek-flash-Abliterated and deepseek-flash-fast; TypeSafe Jev (System One). The GLM, Kimi and DeepSeek models run on private deployments.

Endpoints

OpenAI and Anthropic compatible Chat Completions, Responses, Messages, image generation and edits, video generation, and /v1/systemone for Jev. Keep your SDK; change the base URL and the key.

Routes and prices

Shared pool · Official line · Auto route Same upstream, different routing and price: for Claude, OpenAI and Grok the shared pool is 50% of list price and the official line 90%. Jev has a fixed price. The GLM, Kimi and DeepSeek models use private deployments with fixed prices. An auto-route key sends each request to the lowest-priced line that serves its model.

Billing

Per token · per image · per second Text bills per million tokens for input, output and cache; images per picture by size; video per second by resolution. Every request writes a usage-log row with the model, the route and the charge.

Content retention

Zero Prompts, completions and uploaded files are relayed in memory and never written to disk. Logs hold token counts, model names, routes and identifiers.

Credit

Prepaid USD Top up by card on Waffo’s checkout, in USDT or USDC on NOWPayments, or with a gift card; or take a monthly plan. The balance is only ever spent by your own requests and purchases, and every deduction is listed.

Where the models come from

Claude, OpenAI, Grok and Jev are the providers' own models. The shared pool and the official line reach the same models; they differ in how the traffic is routed and in price. Neither route substitutes a distilled or third-party copy.

The GLM, Kimi and DeepSeek lines are served from private deployments, and every model on them is labelled that way. GLM-5.3-Abliterated, GLM-5.3-Flash-Abliterated and deepseek-flash-Abliterated have the refusal alignment removed; deepseek-flash-fast is DeepSeek V4.1 Flash 1017 on standard weights, and Kimi K3.1 Pro is Moonshot's model. GLM-5.3-Abliterated accepts text only; the other four accept text and images, with 1,048,576 tokens of context, and deepseek-flash-fast writes up to 262,144 output tokens. GLM-5.3-Abliterated, deepseek-flash-Abliterated and Kimi K3.1 Pro also keep a costlier standby deployment: in the rare case that the main one fails before answering, a request can move to it and is billed at the backup price shown on the price table.

How it works

1. Sign in, then add credit on the wallet page.
2. Create an API key. Pick a route, or choose Auto route and let one key cover every model.
3. Point your SDK at https://api.topxai.xyz/v1 and use the key as usual.
4. Follow every charge in the usage log, filtered by key, model or route.

Questions

Do I need a separate key for each model?

No. One auto-route key covers every model and picks the lowest-priced line for each request. Per-route keys are there for when you want to pin a line.

Which price applies to a request?

The fixed USD price of the model on the route the key uses, as listed on the models table before you create the key. The usage log shows the actual charge afterwards.

Is my data stored or used for training?

TopxAI relays prompts, completions and uploaded files in memory and does not persist them or use them for training. Account, billing and usage metadata are retained as described in the Privacy Policy. The upstream provider applies its own retention and training policies.

What are the Abliterated models?

GLM-5.3-Abliterated, GLM-5.3-Flash-Abliterated and deepseek-flash-Abliterated are versions of GLM-5.3, GLM-5.3-Flash and DeepSeek V4.1 Flash 1017 with the refusal alignment removed, served from private deployments at fixed prices, without peak-hour pricing. GLM-5.3-Abliterated is text only; the other two accept text and images. deepseek-flash-fast is not one of them: it keeps DeepSeek's standard weights. All follow the same Acceptable Use Policy as every other model.

How is Jev different from the other models?

Jev is TypeSafe's System One model. It is reached through /v1/systemone or the one-line installer shown under a Jev key, bills input tokens only, and returns output free of charge. The blog has the setup guide.

Who runs it

The TopxAI team runs the service. Support, billing questions and abuse reports: the help button at the bottom right, or support@topxai.xyz.

Models & prices · Docs · Blog · Privacy Policy · User Agreement · Acceptable Use Policy