What Argmax Intelligence Inc. does

Argmax Intelligence Inc. is a Palo Alto-based on-device AI platform company founded in November 2023 by three former Apple ML engineers — Atila Orhon (CEO), Zach Nagengast (CTO), and Brian Keene (Head of Performance) — who previously led ML research and architected Apple's on-device inference engine. The company builds an open-core SDK platform that enables developers to deploy foundation models (speech-to-text, speaker diarization, text-to-speech, and diffusion-based image generation) directly on end-user devices rather than in the cloud, targeting Apple Silicon Neural Engine (iOS/macOS) and Android NPUs (Qualcomm, Google Tensor, MediaTek) via Google LiteRT. Core products include WhisperKit Pro (speech-to-text using Nvidia Parakeet and OpenAI Whisper), SpeakerKit Pro (on-device speaker diarization), DiffusionKit Pro (on-device image generation), Argmax Local Server (Deepgram-compatible WebSocket API for non-native apps), and an Argmax Marketplace for proprietary third-party models.

The company operates an open-source-first go-to-market: WhisperKit, SpeakerKit, and TTSKit are released under MIT license on GitHub (6.8K followers) to build developer community, with a 14-day free trial driving conversion to the paid Pro SDK at $0.42 per device per month (up to 744 hours of audio) — a fixed-price model that the company cites as 744x more cost-efficient than usage-based cloud APIs charging $0.46 per hour. An Enterprise tier adds custom SLAs on latency, accuracy, and battery life, dedicated Slack support, and hardware lab validation, sold via direct enterprise sales. Customers include ModMed (AI medical scribe serving 40,000+ US providers in production since March 2025), Heidi Health, Superwhisper, Macwhisper, and Detail (Apple iPad App of the Year 2025).

Argmax has raised $8M in seed funding (November 2024) led by Salesforce Ventures with participation from General Catalyst and notable angel investors. The company maintains technology partnerships with Nvidia (Parakeet model optimization), Google (LiteRT for Android), Qualcomm (Snapdragon NPU), pyannoteAI (Marketplace anchor), and Hugging Face (model hosting). Reported customer outcomes include 567% MAU growth within Superwhisper, 2x Free Trial to Pro conversion rate improvement, sub-500ms p95 transcript finalization latency for ModMed, and best-in-market 160ms real-time transcription latency on iPhone 12.

Name

Argmax Intelligence Inc.

Legal name

Argmax, Inc.

Website

https://argmaxinc.com

Company type

Private

Founded year

2023

Operating status

Operating

Short description

Argmax Intelligence Inc. builds an on-device AI inference platform that lets developers deploy foundation models — including speech-to-text, speaker diarization, and image generation — directly on user devices across iOS, macOS, and Android via optimized open-source and commercial SDKs.

Ownership category

akta.pro rank

Product category

On-Device AI Inference SDK

NAICS

Software Publishers (5132), Software Publishers (513210), Computer Systems Design and Related Services (54151), Computer Systems Design Services (541512), Custom Computer Programming Services (541511)

SIC

Services-Prepackaged Software (7372), Services-Computer Programming Services (7371), Services-Computer Integrated Systems Design (7373)

akta.pro primary industry

On-Device Inference Runtimes & SDKs (mobile/embedded) (HDAAAJAB)

akta.pro secondary industries

Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem) (HDAEANAC), Developer Tooling & SDK Platforms (BPAMAOAD), AI Application Enablement Platforms (Copilot/Agent Frameworks, SDKs) (HDAEANAJ)

Keywords

• On-device AI SDK
• Speech-to-text inference
• Speaker diarization
• Edge AI deployment
• Foundation model runtime

Headquarters

HQ city

Palo Alto

HQ country

United States

HQ region

North America

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Marketing or Sales, Operations, Infrastructure

Revenue model

1. Argmax Pro SDK Subscription: Per-device monthly/annual subscription. $0.42 per month per device (up to 744 hours/month) vs cloud API pricing of $0.46 per hour. Three tiers: Open-source (free), Pro (paid), Enterprise (custom SLA). Upgrades from open-source to Pro via 14-day trial on the platform. Revenue is predictable per device with unlimited usage, contrasting with usage-based cloud billing that is rate-limited and concurrency-limited.
2. Enterprise SDK Licensing: Enterprise tier with custom pricing, dedicated Slack support, and SLAs on customer-selected metrics (end-of-session latency, accuracy, battery life). Includes hardware lab validation against production device profiles and proactive iOS Developer Beta testing. Sold via direct enterprise sales.
3. Argmax Local Server: Server product for non-native apps (Electron, Python, Node) that cannot integrate native SDKs. API-compatible with Deepgram, enabling migration by changing the host URL. Sold as part of Pro/Enterprise subscription.

Pricing tiers

Go-to-market motion4 records

Distribution channels6 records

Marketing channels10 records

Core offering

Argmax sells a commercial on-device AI SDK (Argmax Pro SDK) that lets developers run frontier foundation models — including speech-to-text (Nvidia Parakeet, OpenAI Whisper), speaker diarization (Pyannote, Sortformer), and image generation (Flux.1, Stable Diffusion) — directly on Apple and Android devices without cloud round-trips. The product combines an open-source SDK family (WhisperKit, SpeakerKit, TTSKit under MIT license) with a paid Pro tier distributed via Swift Package Manager and Maven, plus Argmax Local Server for non-native apps and an Enterprise SDK with SLAs for healthcare and enterprise deployments.

Product overview

Argmax Intelligence Inc. is an on-device AI platform company offering Foundation Models On Device (FMOD). The product portfolio consists of Argmax SDK (open-core architecture with open-source and Pro tiers), Argmax Local Server (for non-native apps), and Argmax Marketplace (proprietary model catalog). The SDK family includes WhisperKit Pro for speech-to-text (Nvidia Parakeet, OpenAI Whisper), SpeakerKit Pro for speaker diarization (Pyannote, Sortformer), and DiffusionKit Pro for image generation. Pro SDK is available for iOS/macOS (Swift) and Android (Kotlin). Enterprise tier offers SLA-backed support with custom metric monitoring.

Differentiator

Problem solved

Functional benefit

Brands

• WhisperKit: Open-source and Pro SDK for on-device Whisper speech-to-text inference on Apple devices
• SpeakerKit
• DiffusionKit
• Argmax Pro SDK
• Argmax Playground
• Argmax Local Server
• Argmax Enterprise SDK

Products and services

• WhisperKit Pro Turn-key on-device speech-to-text SDK supporting Nvidia Parakeet and OpenAI Whisper models with real-time streaming, word timestamps, custom vocabulary, and SRT/VTT subtitle export. Designed for developers building voice features on iOS, macOS, and Android without cloud round-trips.
• SpeakerKit Pro On-device speaker diarization SDK identifying who spoke when in audio. Supports Pyannote and Nvidia Sortformer engines, with ~10MB footprint and the ability to diarize 4 minutes of audio in ~1 second on iPhone.
• DiffusionKit Pro SDK for running advanced diffusion-based image generation models on Apple Silicon devices via CoreML and MLX. Supports Flux.1 and Stable Diffusion model families.
• Argmax Pro SDK Commercial on-device AI SDK bundle combining WhisperKit Pro and SpeakerKit Pro with frontier accuracy, 160ms real-time transcription, custom vocabulary, and per-device subscription pricing. Available for iOS, macOS, and Android.
• Argmax Pro SDK for Android Kotlin-first version of the Argmax Pro SDK for Android, built on Google LiteRT with NPU acceleration across Qualcomm Snapdragon, Google Tensor, and MediaTek chips. Distributed via Maven and the Google Play On-device AI delivery channel.
• Argmax Enterprise SDK Enterprise tier of the Argmax SDK with custom pricing, SLAs on customer-selected metrics (latency, accuracy, battery life), dedicated Slack support, hardware lab validation, proactive iOS Developer Beta testing, and individualized model optimization.
• Argmax Local Server Server product that brings on-device AI inference to non-native apps (Electron, Tauri, Python, Node, Rust) without SDK integration. Exposes a WebSocket API compatible with Deepgram Streaming Speech-to-text. Available via Node.js and Python client packages.
• WhisperKit (Open-source) Open-source Whisper inference framework under MIT license, providing a Swift package for Apple-platform Whisper inference with sample iOS app and Python optimization tools. The free community tier that drives Argmax Pro SDK adoption.
• TTSKit Open-source text-to-speech framework for on-device long-form and real-time speech generation, distributed as part of the Argmax open-source SDK family alongside WhisperKit and SpeakerKit.
• Argmax Playground Demo application for testing Argmax SDK speech recognition, speaker diarization, and custom vocabulary capabilities without writing code. Available on iOS via TestFlight and Android via Google Play Store.

Quantifiable outcome

• 160ms real-time transcription latency vs ~640ms for top cloud providers at equivalent accuracy
• +8 more outcomes

Named customers5 records

Segments4 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration8 records

AI capability6 records

Feature8 records

Partnerships

Four partnerships are on record, tiered core.

• Google LiteRTcoreTechnology or Integration · 18 March 2026Google LiteRT (successor to TFLite) is the inference runtime for Argmax Pro SDK for Android. LiteRT consolidated NPU support across Qualcomm Snapdragon, Google Tensor, and MediaTek chips. LiteRT's ahead-of-time (AOT) compilation addressed the single biggest technical limitation of prior TFLite-based approach (costly JIT compilation). Argmax became the first commercial SDK to ship with LiteRT supporting all major NPU vendors in the Android market. Collaboration documented in a Google developer blog post.
• pyannoteAIcoreTechnology or Integration · 23 June 2025Industry-leading speaker recognition technology company. pyannoteAI's flagship commercial model for speaker diarization is now available on Argmax Marketplace as a SpeakerKit-compatible model upgrade. No code changes required for existing customers to upgrade. Argmax's Interspeech 2025 paper benchmarks show pyannoteAI's commercial model achieves highest diarization accuracy among competitors, with Argmax SpeakerKit achieving second highest while being ~10x faster. pyannoteAI's models remain industry-leading in speaker recognition accuracy.
• Qualcomm AI HubcoreTechnology or Integration · 22 October 2024First Android foray in collaboration with Qualcomm. Deployed OpenAI Whisper on Qualcomm Snapdragon NPUs for speech-to-text on pre-recorded audio, built on Google TFLite (predecessor to LiteRT). WhisperKit Android was the first outcome of this collaboration, enabling best-in-class NPU performance on Samsung Galaxy and other Snapdragon devices.
• NvidiacoreTechnology or IntegrationArgmax uses Nvidia's open-source Parakeet speech-to-text models (Parakeet v2, v3) as the primary models in its Pro SDK. Parakeet v2 ranked #1 on OpenASR leaderboard as of May 2025. Argmax reimplemented and optimized Parakeet for Apple Silicon Neural Engine and Android LiteRT, achieving best-in-market accuracy and speed. Nvidia Parakeet v3 was released one day before Interspeech 2025 and was already running on-device via Argmax SDK at the conference expo.

Scale indicators10 records

Recent moves6 records

Expansion highlights6 records

Broad incumbents

• Apple: Platform owner whose SpeechAnalyzer framework (introduced at WWDC 2025) competes directly with Argmax on Apple devices for on-device ASR. Apple bundles access to its frameworks with iOS/macOS, making it a broad incumbent threat rather than a direct peer product.
• Google Cloud Speech-to-Text: Cloud speech recognition service that competes with Argmax for transcription workloads while also providing Google LiteRT, the inference runtime Argmax Pro SDK for Android is built on. Operates as both a competitor and an underlying infrastructure enabler.
• Microsoft Azure Speech: Cloud-based speech-to-text and speech services within Microsoft Azure AI, competing with Argmax for enterprise transcription workloads at scale across industries including healthcare. Offers both real-time and batch APIs.

Emerging players

• OpenAI: Creator of the Whisper speech recognition models that Argmax WhisperKit originally built on. OpenAI's hosted Whisper API competes in cloud STT while Argmax's value-add is running Whisper and successor models (Parakeet) on-device at frontier accuracy.
• pyannoteAI: Industry-leading speaker diarization technology company whose commercial model is distributed via Argmax Marketplace. Functions both as a partner (model distribution) and as an adjacent competitor in speaker-recognition accuracy benchmarks.

Direct peers

• Speechmatics: Cloud and on-premise speech recognition provider offering ASR for enterprise and developer use cases. Competes with Argmax on accuracy benchmarks and provides alternative deployment options for production transcription workloads.
• Deepgram: Cloud-based speech-to-text provider with a streaming API that Argmax Local Server is explicitly designed to be API-compatible with, enabling zero-code migration. Both compete for the same speech-recognition workload in developer-facing apps, differing only on deployment model (cloud vs on-device).
• Picovoice: On-device voice AI SDK for wake-word, ASR, and voice-controlled interfaces on mobile and embedded. Closest direct competitor to Argmax in the on-device inference SDK category for speech, with similar per-device licensing and platform coverage.
• AssemblyAI: Cloud-based speech-to-text and audio intelligence API competing for the same developer use cases as Argmax Pro SDK. Argmax's benchmarks explicitly compare keyword recognition against AssemblyAI and position on-device as a cost/latency alternative.

Others

• Hugging Face: Model hub and MLOps platform that hosts Argmax's model repositories (argmaxinc/whisperkit-pro, speakerkit-pro, parakeetkit-litert-pro) and serves as distribution infrastructure. Functions as an ecosystem enabler and adjacent AI tooling rather than a direct competitor.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks7 records

Key highlights7 records

Customer concentration

• Xx.com/argmax
• GitHubgithub.com/argmaxinc
• LinkedInlinkedin.com/company/argmaxinc

Compliance2 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles3 records

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Argmax Intelligence Inc.