The KNEO 330 is Kneron’s latest and second private Edge GPT server, following its first unit, the KNEO 300, which launched in 2023 and already has enterprise customers in manufacturing, financial services, and universities, including Stanford and UCLA. The KNEO 330 boasts 48 TOPs of AI computing power, up to 8 concurrent connections, and supports LLMs and stable diffusion. Its RAG accuracy is impressively similar to cloud solutions under much lower hardware conditions. The KNEO 330 drastically reduces the overall cost of AI for small enterprises by 30%-40%.

Compared to most cloud solutions, the KNEO 330 enables easier integration and hierarchical permission management, radically improving privacy and security. Its comprehensive capabilities include offline versions of multimodal GPT.

“AI has seen a recent boom and powerful AI models are advancing more rapidly than many experts thought possible. Balancing ethical AI with profit-driven AI is a real challenge. Concerns such as the required power and data for training the AI models and the potential for AI hallucinations are real problems. We believe our products are key to solving the current GPT ESG & power consumption issues.” says Albert Liu, CEO and founder of Kneron.

In addition to the KNEO 330, Kneron is releasing its AI-embedded PC featuring its 3rd generation NPU chip, the KL830. The AI PC era is here, with sales set to grow from 50 million units in 2024 to over 167 million by 2027, accounting for over 60% of the overall PC market, according to an IDC report. Meanwhile, Gartner predicts the global shipments of AI PCs and AI Smartphones will reach 295 million units in 2024, more than ten times the figure of 29 million for 2023. 

The power and cost of the KL830 allow for a lower-cost AI PC that will enable major accessibility and adoption from a broader market of consumers. The KL830 provides consolidated calculation power (CCP) up to 10eTOPS@8bit with a peak power consumption of 2W. The NPU makes personalized GPT possible.

Demonstrating the power and efficiency of the NPU, when combined with a leading GPU, it saved 30% on energy use and extended product lifetime. This showcases its future in high-powered but accessible gaming PCs.

This chip is ready to be used in AIoT devices. The KL830 chip allows the fixed point to remain consistent with floating point accuracy. The KL830 is also available via a USB dongle that enables any device, whether a broadband router, IoT camera, or classic PC, to become Edge AI-enabled. It provides 10eTOPS@8bit and supports many large language models with lower parameters.

Designed for developers, the KNEO platform is an easy-to-use open marketplace for Edge GPTs. Kneron enables user-friendly AI-generated content model deployment through Kneron’s compiler with a ‘Hugging Face’ link. Users can switch Edge GPTs as required.

Comprised of the developer platform, management platform, and Edge GPT warehouse, Kneron’s Edge GPT as a Service (EGaaS) enables devices to process data locally, allowing real-time decision-making without relying on the Internet or cloud servers. This approach boosts speed, supports multi-modal capability, and enhances privacy and security. Kneron’s comprehensive enterprise-edge GPT solutions can be customized to meet each enterprise's needs by training and deploying large language models (LLMs) for various scenarios.

Kneron has raised nearly $200 million from Horizons Ventures, Qualcomm, Sequoia, Foxconn, Delta, Vivotek, and more. Kneron provides end-to-end integrated hardware and software solutions that enable on-device edge AI inferencing. Qualcomm, Toyota, Kenwood, Garmin, Panasonic, Quanta, Compal, Unimicro, MiTAC, Hanwha, Spark, Naver, TUL, and Gree have adopted Kneron solutions.