AI model developer Motif Technologies will launch a global application programming interface (API) business based on a next-generation model it plans to release next year. It aims to drive monetisation mainly in overseas coding and AI agent markets, where use of open-weight models is relatively active.

On Nov. 11, the industry said Motif has begun developing its next AI model, Motif 4. It aims to release the model in the first quarter of next year in an open-weight format and rank in the top five in global AI model performance evaluations. It has also completed a bridge-round funding raise for the development, and Nvidia is expected to participate alongside existing investors.

Motif plans to step up its API business from next year based on the next model and generate meaningful revenue. It has registered its in-house model on the Microsoft Azure Marketplace, but it has not started a business that runs its own inference infrastructure and supplies APIs.

Motif sees higher commercialisation potential in global markets than in South Korea. Overseas, a market has formed in which various open-weight models are run on in-house infrastructure and supplied through APIs. In South Korea, it sees use concentrated on a few AI services such as ChatGPT and Claude, making it relatively difficult for new models to enter the market.

Motif CEO Jeong-hwan Lim (임정환) recently met with a reporter and said, "Korea is a strange market that only uses GPT and Claude. Even if open-source models come out, there are not many attempts to actually try them." He added, "In the U.S., engineers try new models as soon as they come out. They look at them directly and try to understand them."

In the enterprise AX market, where it initially envisioned commercialisation, Motif sees supplying lower-cost enterprise plans as a long-term alternative to building in-house systems. It believes that as South Korean companies fear the costs of adopting AI and the responsibility for failure, a model of company-wide adoption of global AI firms' services will take hold.

In the global API business, it expects model performance and the cost of generating tokens to determine profitability. It says it needs both performance that users are willing to pay for and lower token production costs to make a profit.

Motif says the basis of its cost competitiveness is its in-house model design technology and GPU optimisation capabilities. In a government standalone AI foundation model project, it proved its capabilities by developing Motif 3 in about five months using 768 GPUs. In August, Motif 3 ranked 10th in the world at the time on the Artificial Analysis Artificial Analysis Intelligence Index (AAII).

It also views its small organisation of about 30 employees and its relatively low valuation as conditions that favour profitability. Unlike global AI companies that must deliver revenue commensurate with large-scale investment and high valuations, it calculates it can secure business sustainability with relatively less revenue.

It has also moved to secure high-quality training data for developing the next model. Motif is contacting major global AI data providers and exploring possible cooperation. It believes capability is important not in buying data itself but in linking it to improvements in model performance.

It is also pursuing cost efficiency in model size. It expects it can develop the next model to be smaller than Chinese models widely used in the market and supply it at a relatively low price.

Its key target areas are coding and AI agents. It believes API demand is large because token consumption is high in the process of repeatedly performing complex tasks. Motif plans to boost coding and agent performance in the next model and secure competitiveness in global markets.

Lim said, "No one uses a model that is vaguely good. We have to make a model that is clearly good, and we have to make a clearly good model cheaply." He added, "I think large language models are not about selling a model but selling tokens made by the model. You have to make tokens that users feel are worth paying for, and the production cost has to be low to leave a profit."

Turning its technical capabilities into actual revenue remains a task. Whether the next model's performance leads to securing actual users and generating revenue will be key.

Lim said, "Right now, the act of making AI itself is considered meaningful, but in a generation where AI becomes everyday infrastructure, the value created on top of it will become important." He added, "The goal is to survive as one of the AI model suppliers until then."