FastMetal and Fixstars Launch High-Speed AI Inference Service on GPU Servers Located in Japan
November 5, 2026 – PSI-Japan K.K. (Kokusai Chotatsu Joho; Headquarters: Chiyoda-ku, Tokyo; hereinafter "PSI-Japan"), operator of the AI gateway "FastMetal," and Fixstars Corporation (TSE Prime: 3687, US Headquarters: Irvine, CA), a leading company in performance engineering technology, today announced the launch of a service, available through FastMetal from November 5, 2026, that provides fast access to the latest AI models running on GPU servers in Japan.
The GPU servers and the AI models offered through the service are managed by Fixstars, whose optimization technology delivers high-speed inference. Using the FastMetal API, users can immediately integrate AI processing into their business operations without their data leaving Japan. As of November 5, 2026, the available model is "DeepSeek-V4.1-Flash-1017," with more models planned.
Service Features
- Input data is processed exclusively on GPU servers owned by Fixstars in Japan
- Inference is optimized with Fixstars technology for fast processing
- Easy integration into existing applications via the FastMetal API
- Available from the day of sign-up by purchasing tokens
Service Overview
- Model ID: japan-deepseek-v4.1-flash-1017
- Launch date: November 5, 2026
- Access: API (endpoint: https://api.fastmetal.ai/v1)
*For pricing, please see the official FastMetal page (https://fastmetal.jp/pricing).
About "DeepSeek-V4.1-Flash-1017"
DeepSeek-V4.1-Flash-1017 is an open-weight model released by DeepSeek on October 17, 2026. Its main specifications are:
- Total parameters: 552 billion (MoE)
- Active parameters: 8 billion on the input side and 16 billion on the output side (based on a new encoder-decoder architecture)
- Context length: up to 1 million tokens
- Input modalities: text and images
*Specifications are based on DeepSeek's official announcement (https://api-docs.deepseek.com/news/news261017/).
Background of the Collaboration
As business use of generative AI expands, the performance of open-weight models (models whose weights are publicly released and which can run in closed-network environments such as on-premises systems) is approaching that of cutting-edge commercial models. At the same time, running these models at practical speed and cost requires software optimization that gets the full performance out of GPUs. In addition, for companies handling business data, where that data is processed has become an important criterion when choosing AI.
Under its corporate message "Speed up your AI," Fixstars has been working to accelerate the inference and training of AI models. FastMetal is an AI gateway that provides access to AI models from Japan and around the world with a single API key. Through this collaboration, models optimized and operated by Fixstars in Japan are now easy to use through FastMetal.
Roles of Each Company
- Fixstars: Building, optimizing, and operating the inference infrastructure on its GPU servers in Japan
- PSI-Japan: API provision, sales, billing, and user support through FastMetal
Comments from the Representatives of Both Companies
Satoshi Miki, President & CEO, Fixstars Corporation
"As the performance of open-weight models rapidly improves, accelerating inference is essential to harnessing their capabilities in business. We are delighted to deliver the latest AI models, optimized and operated by Fixstars in Japan, to more developers and companies through FastMetal. We will continue to support the adoption of generative AI in Japan with our acceleration technology."
Darshaun Nadeau, Representative Director, PSI-Japan K.K.
"We are very encouraged to collaborate with Fixstars and to make the models it optimizes and operates in Japan available to FastMetal customers. FastMetal's core philosophy is 'Your data, your choice.' We believe it is important that customers themselves can choose whether their data may be used for training, whether it may be stored, and in which country's servers it is processed. For customers who want their data processed in Japan, this model is a compelling option. Together with Fixstars, we will continue to expand those options."
Future Plans
Going forward, the two companies plan to progressively expand the lineup of models on FastMetal that Fixstars optimizes and operates in Japan.
About Fixstars Corporation
Fixstars is a technology company whose corporate message is "Speed up your AI." Leveraging software optimization technology that makes the most of computing resources, Fixstars dramatically accelerates both AI model inference and training. The company advances next-generation AI technology across a wide range of fields, including healthcare, manufacturing, finance, and mobility.
- Headquarters: 31F BLUE FRONT SHIBAURA TOWER S, 1-1-1 Shibaura, Minato-ku, Tokyo 105-0023, Japan
- Representative: Satoshi Miki, President & CEO
- Founded: August 2002
- URL: https://www.fixstars.com
About PSI-Japan K.K.
Founded in 1981, PSI-Japan provides corporate customers with management services for internet assets, including domain names. The company operates FastMetal, an AI gateway that provides access to LLMs and image and video generation models through a single API. With FastMetal, customers choose whether their data may be used for training, whether it may be stored, and in which country's servers it is processed, and requests are automatically routed to match those conditions.
- Headquarters: Asahi Sanbancho Plaza 208, 7-1 Sanbancho, Chiyoda-ku, Tokyo 102-0075, Japan
- Representative: Darshaun Nadeau, Representative Director
- Founded: May 1981
- URL: https://fastmetal.jp
*Company names, product names, and service names mentioned in this document are trademarks or registered trademarks of their respective owners.
Media Contact
Fixstars Corporation, Public Relations
Email: press@fixstars.com
Tel: +81-3-6420-0751
PSI-Japan K.K., FastMetal Team
Email: team@fastmetal.ai