A strategic shift appears to be underway at OpenAI, which has traditionally focused on closed artificial intelligence (AI) models. OpenAI has announced plans to release an open-weight model—the first such move since ChatGPT-2 nearly six years ago.
According to the ICT industry on Monday, OpenAI is expected to launch a reasoning-based open-weight model in the near future. Sam Altman, CEO of OpenAI, recently shared this news via X (formerly Twitter).
An open-weight model sits between closed and open-source models. While open-source models release all code and data, open-weight models disclose information such as which data are prioritized (weights) and partial code—essentially a semi-open model.
Altman said it was something he had been considering for a long time and added that he would collaborate with developers to make it as useful as possible.
OpenAI’s decision to enter the open-model race appears to be influenced by the growing popularity of models like China’s DeepSeek. As a cost-effective, high-performance open model, DeepSeek has attracted global attention and a wide user base. Unlike closed models, open models allow users to modify the system according to their specific needs.
Following the impact of DeepSeek, Altman hinted at a strategic shift in January, stating that he felt OpenAI may have been on the wrong side of history regarding open source and suggested the need for a new approach.
OpenAI’s entry into the open-model market seems to be a move to attract a broader user base and counter competitors such as DeepSeek, Meta, and Google.
An industry insider commented that OpenAI has long been a symbol of closed models, so releasing an open-weight model appears to signal a move toward openness. Open models can draw in large user communities, and OpenAI likely adopted this strategy to avoid losing users to competitors.
With OpenAI joining the competition, the battle over open models is expected to intensify. In South Korea, LG also open-sourced its reasoning-based AI model EXAONE Deep this past March.