Desert Ant Labs
Monthly visits
47,637
2026-09

Desert Ant Labs - On-device AI models for app features without per-call costs

Desert Ant Labs provides small AI models and a native SDK for adding audio, text, and vision features to apps. Models run on-device, and the homepage says inference is unlimited with each model free up to 100,000 monthly active devices. Available model pages include PII redaction and structured JSON extraction; Schemer is in closed beta.

Visit WebsiteUpdate this product

0Views

Capabilities

On-device AI models

Run models for audio, text, and vision tasks on a user's device. The catalog includes speech recognition, language detection, image classification, content moderation, and text processing models.

PII redaction

Redact personal data in text before it reaches a server, using the Redact model in a browser, on a phone, or in a Node backend. It can also restore masked values after an external model response.

Structured JSON extraction

Schemer extracts typed fields from text against a supplied JSON Schema and returns structured JSON. Missing fields can remain empty; the model is in closed beta, with early access available on request.

PricingFreemium

The homepage states that models are free up to 100,000 monthly active devices, with unlimited inference. Compare the monthly active device threshold and model availability; contact the company for licensing above the threshold.

PlanPriceIncluded features and limits
Free

Free up to 100,000 monthly active devices

  • Inference is unlimited.

  • The allowance applies separately to each model. Such pricing is stated on the homepage.

  • Licensing above 100,000 monthly active devices requires contacting the company.

Use cases

Add on-device AI features to an app

When an app needs audio, text, or vision processing without sending inputs to a cloud model, developers can use the native SDK and select a model for the task. The model runs on-device, and inference is unlimited under the stated free allowance.

Redact personal data before submission

When a user enters personal information in a form or support message, developers can run Redact in the browser or on a phone before submission. The text is masked before it reaches the backend, and masked values can be restored in supported round trips.

Limitations

  • Each model is free only up to 100,000 monthly active devices; contact the company about licensing above that threshold.

  • Schemer is in closed beta and requires an early-access request; its availability differs from the generally available models.

Who it is for

  • App developers adding on-device audio, text, or vision features

  • Developers who need to mask personal data before sending text to a backend or external model

FAQ

How much does using a model cost?

Each model is free up to 100,000 monthly active devices, and inference is unlimited within that allowance. Contact Desert Ant Labs about a license above the threshold.

Where do the models run?

The models are designed to run on-device. The Redact page also describes browser and Node backend use, and the company lists Swift, Kotlin, and JavaScript platforms.

Is Schemer generally available?

No. Schemer is in closed beta, and early access is available by request. Its model card and weights are public.

Comments

Please sign in to leave a comment.
No comments yet. Be the first to share your thoughts!