.jpg)
Fujitsu Limited announced on September 7, 2025, a groundbreaking generative AI reconstruction technology for its Takane Large Language Model (LLM), a core component of the Fujitsu Kozuchi AI service. This innovation enables the creation of lightweight, energy-efficient AI models, significantly advancing performance and sustainability.
Fujitsu’s new technology leverages two key innovations: 1-bit quantization and specialized AI distillation. The proprietary 1-bit quantization applied to Takane reduces memory consumption by 94% while retaining 89% accuracy compared to unquantized models, achieving a three-fold increase in inference speed. This allows large generative AI models, previously reliant on multiple high-end GPUs, to run efficiently on a single low-end GPU.
The world-first specialized AI distillation, inspired by brain processes, not only shrinks model size but also enhances accuracy beyond the original model by extracting and condensing task-specific knowledge. This creates highly efficient, reliable AI systems tailored for specific applications.
This lightweighting capability democratizes advanced AI, enabling deployment on edge devices like smartphones and factory machinery. Benefits include:
For example, in sales negotiation prediction using Fujitsu’s CRM data, the technology achieved an 11x inference speed increase, 43% accuracy improvement, and 70% reduction in GPU memory and operational costs. In image recognition, it improved unseen object detection accuracy by 10% over existing methods.
Fujitsu plans to roll out trial environments for Takane with this technology in the second half of fiscal year 2025, alongside releasing quantized versions of Cohere’s open-weight Command A models via Hugging Face. This aligns with Fujitsu’s broader strategy to enhance its Kozuchi AI service and support enterprise digital transformation.
The technology’s ability to operate on low-end hardware could transform industries like healthcare, manufacturing, and finance, where secure, real-time AI is critical. Fujitsu’s collaboration with Cohere and Nutanix further amplifies Takane’s reach, with deployments planned for hybrid cloud environments.
As AI adoption grows, energy efficiency and edge computing are critical to reducing costs and environmental impact. Fujitsu’s advancements address these challenges, positioning Takane as a leader in sustainable AI. The company’s stock (FJTSY) is priced at $20.98 as of September 8, 2025, with a market cap of $39.59 billion, reflecting a 2.39% intraday gain, signaling investor optimism in its AI innovations.
Fujitsu Limited is a global leader in digital transformation, delivering innovative AI, computing, and sustainability solutions. With over 124,000 employees across 100 countries, Fujitsu’s Kozuchi AI service and Takane LLM empower enterprises in industries like finance, manufacturing, and healthcare. Fujitsu is committed to solving societal challenges through cutting-edge technology.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.