The Pulse
DMM and Algomatic Dynamics launch router for 60 or more AI models
DMM AI Router connects businesses to generative AI models through a single API and can switch models when errors or delays occur. The companies plan to add image and video support, budget limits and access controls.

DMM AI Router gives businesses access to 60 or more generative AI models through a single API, including GPT, Gemini and Kimi models. DMM.com and DMM group startup Algomatic Dynamics jointly developed the gateway, according to their November 19 release.
The service lets customers compare models and set rules for choosing them by task. It can also switch automatically to another model when the one in use encounters an error or delay.
One connection for multiple models
The gateway consolidates access to models that would otherwise involve separate provider contracts, development against different API specifications and API key management, the companies said. Businesses can test different models over the same connection and compare answer quality, response speed and cost.
The product is compatible with OpenAI's Chat Completions API, according to its service page. Users specify the router endpoint and a model ID to make requests to supported models.
There is no initial or monthly fee, but payment is by prepaid credit. Fees apply when credit is added, and consumption tax is charged separately. Users can check request counts, usage charges and credit balances on their account page.
Switching by task or failure
Customers can set rules for using different models for different jobs. The release gives the example of using a lower-cost model for frequently asked questions and a higher-performance model to analyze important documents.
During a proof of concept, users can manually switch among models for comparison. When moving to production, they can use the same connection with rule-based switching and automatic switching if a model encounters an error or delay.
Features planned next
DMM and Algomatic Dynamics plan to add support for models that handle images and video, alongside text. Planned capabilities include reading images, summarizing video and generating images from text. They also plan to increase the number of available providers and models.
A planned guardrail feature would let customers specify permitted models and set budget ceilings. Requests that do not meet those conditions would be restricted.
Another planned feature would separate administrator and user permissions, including who can issue and manage API keys or change usage rules. These controls are among the additions the companies intend to introduce over time.
Sources
- Original source · Press release DMM.com via PR TIMES prtimes.jp
Other pages read
- DMM AI Router airouter.dmm.com