- Specialization
Speechand 1 more specialization valuesLess
- Language Models
- Business
Model Developmentand 1 more business valuesLess
- AI Research
- Market
Software Developmentand 1 more market valuesLess
- Manufacturing
- Ecosystem
- Not classified
- Customers
Developersand 1 more customers valuesLess
- Enterprises
More details
- Organization type
- Commercial Company
- Ecosystem
- Not classified
- Business
- Model Development, AI Research
- Specialization
- Speech, Language Models
- Market
- Software Development, Manufacturing
- Customers
- Developers, Enterprises
- Delivery models
- Open Source, On Premise Software, Professional Services
- Revenue models
- Licensing, Services Fees
- Listing status
- Not classified
- Latest equity round
- Seed
- Reported funding amount
- Not classified
About
Stellon Labs is an AI research lab building tiny frontier models for speech, language, and video that run on smartphones, wearables, embedded systems, and other edge devices. It targets developers and software and hardware companies seeking local AI, differentiating through compact, CPU-capable models such as the open-source KittenTTS, which is available in versions as small as 25 MB.
Market
Stellon Labs competes in edge AI and on-device speech intelligence, positioning itself around frontier-quality models that are small enough to run on minimal hardware. Its differentiation is the combination of open-source distribution, very small model sizes, ONNX and CPU-based inference, and support for offline, privacy-sensitive deployments, while competitors such as Picovoice, Kyutai, and Piper also provide local or on-device speech capabilities.
Stellon Labs targets developers and hardware manufacturers building AI features for resource-constrained devices such as smartphones, wearables, embedded systems, and browsers. The likely buyers are engineering, product, and embedded-AI teams that need private, low-latency, offline inference without GPUs.
At a Glance
Problem
Stellon Labs addresses the mismatch between modern foundation models and the limited compute, memory, power, and connectivity of edge hardware. Large models are difficult to run on smartphones, wearables, robots, laptops, and embedded systems, while cloud inference introduces network latency, privacy exposure, and recurring compute, storage, and orchestration costs. The clearest initial use case is private, real-time voice intelligence—especially text-to-speech running locally on devices without a GPU or constant internet connection.
Product / Service
The company is building tiny frontier models for speech, language, and video, beginning with KittenTTS, an open-source ONNX-based text-to-speech library. Its available models range from 15 million to 80 million parameters and 25–80 MB on disk; the smallest quantized version is 25 MB. It runs on CPUs, offers eight voices and 24 kHz output, and is designed for edge deployment without a GPU.
The apparent delivery model combines open-source models and developer packages with paid integration services. Stellon Labs offers commercial support, custom voice development, enterprise licensing, and is planning an optimized inference engine and mobile SDK. The benefit is to give product and hardware teams capable local AI that is smaller, cheaper to operate, lower-latency, more private, and usable offline than a cloud-dependent alternative.
Market
Stellon Labs competes in on-device AI, edge AI, and tinyML, specifically the model and tooling layer for resource-constrained devices. Its direct or adjacent competitors identified in the research include Deeplite, Genuix, OneBit AI, SynthInt, and Foveal Vision; Picovoice is a notable adjacent competitor in production on-device voice AI. Stellon Labs differentiates through very small frontier models and an open-source developer-led distribution strategy, starting with speech and expanding toward language and video.
The company appears to be an early-stage, developer-led venture rather than a scaled commercial business. YC reports that KittenTTS reached 8,000 GitHub stars and 45,000 model downloads within two weeks of launch, while the repository later showed 15,234 stars and 874 forks, providing meaningful evidence of developer interest. Stellon Labs was founded in 2025 and entered YC’s Summer 2025 batch; Tracxn listed it as unfunded as of July 2026, although a separate third-party source reported $220K of 2025 revenue. Because the revenue figure is not corroborated by a primary source, its commercial status is best described as early commercialization with strong open-source traction, not definitively pre-revenue or proven at scale.
Founders & Leadership
Funding History
Y Combinator
Recent News
468 Capital published a dated company page listing Stellon Labs Inc. The page does not disclose investment size or deal terms.
Active Roles
0No active roles right now.
Get notified when they postBusiness Model
Stellon Labs appears to use an open-source-plus-commercial-services model: it distributes KittenTTS openly while selling commercial support for integration, custom voices, enterprise licensing, and broader custom-model training and deployment work. The evidence does not show a public subscription or usage-based pricing schedule.