OpenAI Reportedly Slows Training and Redirects Resources to Safety
OpenAI is reportedly slowing model training and reallocating human resources to safety, while 80–90% of its research focuses on GPT-7, GPT-8 and future models.
OpenAI Reportedly Slows Training and Redirects Resources to Safety
OpenAI is reportedly slowing down model training and shifting human resources toward safety work.
Boris Power, head of Applied Research at OpenAI, stated that “80–90% of the company’s research is currently devoted to GPT-7, GPT-8 and the next models.”
According to him, improvements within the current generation, including post-training, are rather a forced measure that accelerates training and slightly boosts today’s models. This approach is not the company’s long-term strategy.
The real vision, he said, is to train enormous, super-powerful next-generation models and then distill them into smaller systems. For example, Luna appears to be a distillation of Astra.
“80–90% of the company’s research is currently devoted to GPT-7, GPT-8 and the next models.”
The statement suggests that OpenAI may be trading short-term model upgrades for a broader focus on safety and the development of next-generation AI systems.
Why it matters
AnalysisIt signals a possible shift from incremental upgrades of current LLMs to safety-focused work and next-generation model development. For developers and the market, it may mean fewer rapid improvements to today’s models and more attention to foundational AI research.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
OpenAI GPT (Flagship Models of the GPT Series)
The primary universal line of large language models from OpenAI (GPT-4, GPT-4o). Optimized for complex text analysis, programming, creativity, and daily intellectual tasks.
Synthetic Data Generation & Pretraining
A technology for generating, automatically filtering, and formally verifying training datasets using artificial intelligence to overcome the shortage of quality human data.
OpenAI publishes Apps SDK UI for building ChatGPT app interfaces
OpenAI has released the source code of Apps SDK UI, a React and Tailwind library with ready-made components for creating interactive interfaces inside ChatGPT.
Part I. AI Agent Architecture: The Foundations
The eleven foundational building blocks of modern autonomous agents: model taxonomy, the agentic runtime loop, advanced prompting, distillation, skills architecture, memory systems, function calling, and FinOps.