Announced details
Complex software development, long-running tasks as an agent, business and development
Median time to first token (TTFT) of 1.127 seconds. The input price is approximately one-twentieth of Z.ai's official API price, and the output price is approximately one-third.
Z.ai is the developer of GLM-5.3. Fixstars quantizes and optimizes the model released by Z.ai and provides it through FIC. Capital and personnel relationships are Not disclosed.
China-based Z.ai, the developer of GLM-5.3, which it released in August 2026 as an open-weight LLM. Other information on its location, representative, capital and year of establishment is Not disclosed.
Domestic GPU servers; all inference servers and storage are located in Japan
Per 1 million tokens: input ¥10.49, output ¥210, cache read ¥10.34
The company does not address Contract Value / Impact on Business Performance.
Prices offered by each provider on OpenRouter and prices for Z.ai's official API (as of 20 November 2026)
| Item | Details |
|---|---|
| Names of Products / Services Covered | Fixstars Inference Cloud(FIC), FIC version of GLM-5.3, FIC Launch Commemorative Campaign |
| Counterparty | Z.ai |
| Counterparty overview | China-based Z.ai, the developer of GLM-5.3, which it released in August 2026 as an open-weight LLM. Other information on its location, representative, capital and year of establishment is Not disclosed. |
| Relationship with the Company | Z.ai is the developer of GLM-5.3. Fixstars quantizes and optimizes the model released by Z.ai and provides it through FIC. Capital and personnel relationships are Not disclosed. |
| Offering price | Per 1 million tokens: input ¥10.49, output ¥210, cache read ¥10.34 |
| Launch date | 20 November 2026 (actual) |
| Applicable area | Complex software development, long-running tasks as an agent, business and development |
| Location | Domestic GPU servers; all inference servers and storage are located in Japan |
| Quantitative Effects | Median time to first token (TTFT) of 1.127 seconds. The input price is approximately one-twentieth of Z.ai's official API price, and the output price is approximately one-third. |
| Source | Prices offered by each provider on OpenRouter and prices for Z.ai's official API (as of 20 November 2026) |
Items this document does not state (8)
These are items the company does not state in this document. We do not leave them blank, nor fill them in from other sources or by inference.
The reasons the company gives
The company's own explanation, not something inferred from the numbers.Fixstars quantized and further optimized the official model from Z.ai using NVFP4, reducing computational and memory requirements while achieving both lower prices and shorter response times.