Argmax provides an on‑device AI platform via its Argmax SDK, which runs inference on a device’s Neural Processing Unit (NPU) to deliver real‑time speech‑to‑text, speaker recognition, and image generation without taxing the CPU or GPU. By keeping model execution local, it ensures data privacy, offline operation, and predictable per‑device pricing, with a free open‑source tier and a Pro tier for unlimited usage.
Funding
$8M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
3OFounders
Product
Problem
Developers of mobile and edge applications need high‑accuracy AI capabilities such as speech‑to‑text, speaker recognition, and image generation, but cloud‑based inference introduces latency, privacy risks, and unpredictable usage costs. Existing on‑device frameworks often rely on CPU/GPU, causing performance conflicts with other apps and limiting real‑time experiences.
Solution
Argmax provides an on‑device AI platform delivered through the Argmax SDK. The SDK runs inference on the device’s Neural Processing Unit (NPU), delivering real‑time performance without taxing the CPU or GPU, so other workloads remain unaffected. By keeping model execution local, user data never leaves the device, ensuring privacy and eliminating the need for constant internet connectivity. Argmax offers a free open‑source tier for basic use and a Pro tier with unlimited per‑device usage under a fixed monthly license, giving developers predictable cost structures. The platform includes ready‑made modules—WhisperKit for speech‑to‑text, SpeakerKit for speaker diarization, and DiffusionKit for image generation—plus features like custom vocabulary and cross‑platform support for iOS and Android via Google LiteRT.
Target Audience
Primary customers are mobile app developers and product teams building real‑time AI features—such as speech transcription, voice assistants, speaker identification, or generative imaging—who require on‑device performance, data privacy, and predictable pricing.
Features
- NPU‑first inference engine that avoids CPU/GPU contention and maintains low latency (≈160 ms for real‑time transcription)
- Open‑source SDK (Basic Plan) with optional upgrade to Pro SDK that adds unlimited usage, advanced models, and enterprise SLAs
- Pre‑integrated AI modules: WhisperKit (state‑of‑the‑art speech recognition), SpeakerKit (on‑device speaker diarization), DiffusionKit (Flux and Stable Diffusion image generation)
- Custom Vocabulary support enabling runtime configuration of up to 1,000 keywords for domain‑specific speech recognition (e.g., medical specialties)
- Cross‑platform Kotlin‑first Android SDK built on Google LiteRT and Apple‑native SDK, covering major NPU vendors (Qualcomm Snapdragon, Google Tensor, MediaTek)
- Offline operation: Basic Plan runs fully offline after model download; Pro Plan requires connectivity only once per 30 days for license renewal
- Security posture with SOC 2 compliance in progress and data minimization by design