大模型
www.aibase.com ·
OpenAI 高管称 80%–90% 研发资源已转向 GPT-7 及后续模型
据 The Decoder 报道，OpenAI 应用研究负责人 Boris Power 在 Fellows Forum 2026 表示，公司已将 80%–90% 的研发资源投入 GPT-7、GPT-8 及后续代际，视其为技术价值的核心来源。他称同代内的增量微调（如 GPT-5.1 到 GPT-5.2）依赖特定训练数据、利于快速迭代，但内部视为短期投入；真正的性能突破来自新基座模型发布，每次大版本跃迁都会促使团队重新校准短期重点。他还强调当前 AI 助手的核心挑战在于引导用户理解并释放技术潜力，而非仅模型质量。上述内容为二手转述，具体比例与内部安排未获独立证实。
AI推荐理由 转述 The Decoder 报道的 OpenAI 应用研究负责人 Boris Power 在 Fellows Forum 2026 的发言，称 80%–90% 研发资源投向 GPT-7、GPT-8 及后续代际，并区分同代微调与新基座模型的价值。信息属二手转述、无一手链接或原始演讲材料，具体比例与内部资源分配无法独立核实，故质量分受限；但涉及头部实验室研发重心与代际策略，对判断模型迭代节奏有参考价值。
62
正文
OpenAI executives reveal: 80% to 90% of R&D resources have been shifted to GPT-7 and subsequent models
AIbase基地
Published inAI News · 2 min read · Sep 28, 2026
5
According to The Decoder, Boris Power, the head of applied research at OpenAI, revealed at the Fellows Forum 2026 that the company has fully allocated 80% to 90% of its R&D resources to GPT-7, GPT-8, and subsequent generations, as they constitute the core source of technological value.
Power pointed out that incremental fine-tuning within the same generation (such as the upgrade from GPT-5.1 to GPT-5.2) relies on specific training data and helps teams achieve rapid iteration, but is considered a short-term investment internally; real performance breakthroughs always come from the release of new base models, and each major version jump will prompt the team to recalibrate their short-term focus.
Regarding the evolution of product interaction, Power emphasized that the core challenge of current AI assistants lies in guiding users to understand and unlock the potential of the technology, rather than just model quality.
OpenAIGPT-7GPT-8AINeologisms
This article is from AIbase Daily
Scan to view
Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you understand technical trends, and learning about innovative AI product applications.
—— Created by the AIbase Daily Team
© Copyright AIbase Base 2024, Click to View Source -