OpenAI CEO Sam Altman revealed plans to launch an open-weight AI model by late 2025, reversing its longstanding closed approach amid competition from Meta and DeepSeek.
OpenAI will release its first open-weight AI model since GPT-2 in late 2025, CEO Sam Altman confirmed via X, responding to industry demands for transparency and customization.
Strategic Shift for OpenAI
In a post on X dated 20 October 2025, Altman stated the upcoming model will enable local deployment and modification, addressing developer frustrations with closed systems. This marks a notable policy reversal for OpenAI, which previously defended closed models as necessary for safety.
Competitive Pressures Mount
The move follows Meta’s September 2025 release of Llama 4, now powering 38% of open-source projects according to StackOverflow data. Chinese firm DeepSeek further intensified competition with its energy-efficient DS-7B-Open model launched last week.
Regulatory Drivers
EU’s AI Transparency Act, enacted 15 October 2025, mandates explainability in healthcare and finance sectors. OpenAI’s new model reportedly meets these requirements, potentially easing compliance for European clients.
Analysts note 72% of enterprises now prioritize adaptable AI tools (Gartner, October 2025), suggesting closed models risk losing market share. However, some experts warn open weights could enable misuse – a concern central to OpenAI’s earlier policies.
Historical Context
OpenAI’s 2019 GPT-2 release sparked debate when the company initially withheld the full model over misuse fears. The subsequent GPT-3 and GPT-4 remained closed, contrasting with Meta’s open-source strategy. In 2023, Llama’s leak forced Meta to embrace openness, ultimately strengthening its developer ecosystem.
The EU’s regulatory push mirrors 2021’s AI Act proposals, but with stricter transparency mandates. Similar requirements in China’s 2024 AI Governance Framework suggest global alignment toward auditable systems, particularly for medical and financial applications.
