
What happened
Fixstars began offering GLM-5.3 on Fixstars Inference Cloud on November 20, using an OpenAI-compatible API; the model is Z.ai's open-weights LLM, quantized to NVIDIA's NVFP4 format and further optimized by Fixstars.
Why it matters
All inference servers and storage stay in Japan and inputs and outputs are neither stored nor used for training, so companies that cannot send source code or internal documents to overseas services can use it.
WHO IT HITSJapanese enterprises with data-residency constraints — particularly software development teams and firms handling internal documents — gain a domestically hosted option for a coding-capable open-weights model.
Summaries like this, every morning.
GLM-5.3 is an open-weights LLM that Z.ai published in August 2026, and Fixstars is positioning its hosted version for complex software development and long-running agent tasks. To make that practical, Fixstars quantized the official model to NVIDIA's NVFP4 format to cut compute and memory use, then applied its own optimization on top. The arrangement keeps inference servers and storage entirely within Japan and avoids storing inputs and outputs or using them for training — a combination aimed at firms that cannot send source code or internal documents abroad. On price, Fixstars cites roughly one-twentieth the input and one-third the output cost of Z.ai's official API, while time to first token is a median 1.127 seconds under stated test conditions. Availability is currently limited to direct contract customers, with FastMetal access to follow.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Oxide Computer Co., which raised $445 million in October to expand rack production, built its business on the…
OpenAI now requires gateways and routers that provide third-party access to its AI models to be reviewed and a…

Anthropic shares traded as low as $83.15 on Friday, down more than 14% on their second day of trading, leaving…

Running the official harness, Claude Opus 5.5 wrote an ARC model with no gradient descent and no forward calls…

A Zenn article by Hiromi lays out Continuous Reasoning Acquisition (CRA), a design for capturing veteran judgm…

Major US tech companies are refocusing generative AI competition from answer quality toward agent capability…
