Company
The operator behind the API.
Arnict hosts open-model inference on infrastructure it operates and exposes it through one familiar OpenAI-compatible API. We publish the model, rates, location, and availability so developers know what they are using. The company is Arnict, Inc..
Operating principles
Why Arnict exists, in four commitments.
These principles define the product. They are also the standard the website holds itself to: product facts over hype, always.
Host the model directly
Arnict operates the model-serving path itself. We are not a broker, marketplace, router, or proxy over third-party inference hosts. When you call Arnict, the inference runs on infrastructure we operate.
Leave no content trail
Prompts and completions are processed for the life of a request and then discarded. They are not stored, used for training, or forwarded to a third-party inference provider.
Use a familiar interface
The API is OpenAI-compatible: bearer auth, the chat completions path, and standard message shapes. Developers keep their existing clients and change the base URL.
Keep the facts explicit
Model, location, pricing, and availability are stated plainly and kept consistent across every surface. No invented benchmarks, logos, testimonials, certifications, or guarantees.
What we will not claim
Restraint is a feature.
The website never invents benchmarks, customer logos, testimonials, certifications, performance guarantees, or model availability. If it is on the site, it must be product truth.
Current product truth
- Kimi K3 is the sole designated launch model, directly hosted on US-based Arnict-operated servers.
- Prompts and completions have Zero Data Retention.
- The API is OpenAI-compatible; examples are illustrative until the endpoint is operational.
- Inference availability is “Coming soon”; the status page is the authority.