OpenAI is quietly developing a mysterious project codenamed "Strawberry," whose capabilities may have reached the level of a Ph.D.
TechFlow Selected TechFlow Selected
OpenAI is quietly developing a mysterious project codenamed "Strawberry," whose capabilities may have reached the level of a Ph.D.
This project aims to achieve in-depth research that has not been possible with AI models until now.
Navigating Web3 tides with focused insightsBy: Samuel Nwite
Translation: MetaverseHub
To maintain its leading edge in the fast-evolving field of artificial intelligence, OpenAI is secretly developing a new AI model codenamed "Strawberry."
This information comes from internal documents reported by Reuters and a source familiar with the matter. The Microsoft-backed startup, best known for its ChatGPT product, is now fully focused on demonstrating advanced reasoning capabilities in its models—a potential major leap forward in AI technology.
01. Inside the "Strawberry" Project
According to a recent internal document seen by Reuters in May, OpenAI teams are deeply engaged in the "Strawberry" project. While the exact timeline of the document remains unclear, it outlines OpenAI’s plans to use "Strawberry" for advanced AI research.
The project is described as “ongoing” and has remained highly confidential even within the company. The goal of "Strawberry" is not just to generate answers, but to enable AI to autonomously and reliably browse the internet and conduct what OpenAI calls “deep research.”
“This is something current AI models cannot achieve,” said the source, highlighting the project's ambition.
When asked about "Strawberry" and details in this report, an OpenAI spokesperson said in a statement: “We want our AI models to see and understand the world the way we do. Ongoing research into new AI capabilities is common practice across the industry, and our shared belief is that over time, these systems will become increasingly capable at reasoning.”
However, the spokesperson did not directly confirm or comment on the "Strawberry" project.
02. From Q to Strawberry: A New Era of Reasoning
Reports indicate that "Strawberry" is the successor to a previous project named Q.
According to two sources, OpenAI internally views Q as a breakthrough due to its ability to answer complex scientific and mathematical questions—surpassing the capabilities of most currently commercialized models.
As Bloomberg reported, during an all-hands meeting earlier this year, OpenAI showcased a research project demonstrating new human-like reasoning abilities.
While Reuters could not verify whether the demonstrated project was "Strawberry," it aligns closely with the company’s ongoing efforts to enhance AI reasoning.
OpenAI CEO Sam Altman emphasized the importance of reasoning in AI, stating earlier this year that “the most important area for progress will revolve around reasoning capability.”
03. The Challenge of AI Reasoning
Researchers believe that improving AI models’ reasoning ability is key to achieving human-level or even superhuman intelligence. While large language models can efficiently summarize texts and write articles, they often fail on commonsense questions and logical tasks, leading to so-called “hallucinations” or generation of incorrect information.
According to AI researchers, reasoning involves the AI’s ability to plan, understand the physical world, and solve multi-step problems.
OpenAI’s "Strawberry" project aims to overcome these challenges through specialized post-training processes. This includes fine-tuning AI models after pre-training them on massive datasets.
According to one insider, "Strawberry" shares similarities with Stanford University’s “Self-Taught Reasoner” (STaR), which allows AI models to iteratively create their own training data—potentially enabling them to reach higher levels of intelligence.
Noah Goodman, one of STaR’s creators and a professor at Stanford, commented: “I think it’s both exciting and scary… If things continue in this direction, we humans have some serious thinking to do.”
04. Long-Horizon Task Planning and Autonomous Research
One ambitious goal of the "Strawberry" project is to perform long-horizon tasks (LHT), requiring AI to plan and execute a sequence of actions over extended periods.
Internal documents show that OpenAI is training and evaluating models on a “deep research” dataset to develop these capabilities.
Although the specific content and duration of the dataset remain undisclosed, the objective is clear: to enable AI to conduct autonomous research with the help of computer-using agents (CUA) and take action based on research findings.
05. A Competitive AI Industry
In advancing AI reasoning, OpenAI is not alone. Major tech companies like Google, Meta, and Microsoft, along with numerous academic labs, are also exploring various techniques to improve AI reasoning.
However, opinions differ on whether large language models can incorporate long-term planning and advanced reasoning into predictions. Yann LeCun, a pioneer of modern AI at Meta, frequently expresses skepticism about the ability of large language models (LLMs) to achieve human-like reasoning.
"Strawberry" represents a crucial part of OpenAI’s strategy to address current limitations of AI models. By developing more advanced reasoning capabilities, OpenAI aims to unlock new possibilities—from scientific discovery to creating novel software applications.
At the same time, the company has been signaling to developers and partners that significantly enhanced reasoning technologies are on the horizon.
The development of "Strawberry" includes post-training methods such as fine-tuning, involving human feedback and iterative learning processes. These techniques aim to refine AI models and improve their performance on specific tasks.
Advances through "Strawberry" could redefine the capabilities of AI and set new benchmarks for what these models can achieve.
While the path forward is full of challenges, the potential rewards are immense—heralding a new era of intelligent, autonomous AI systems.
In the words of an OpenAI spokesperson: “We want our AI models to see and understand the world the way we do. If the ‘Strawberry’ project succeeds, we’ll be one step closer to realizing this vision.”
OpenAI has introduced a five-level system to track its progress toward achieving Artificial General Intelligence (AGI). These levels range from Level 1, representing today’s conversational AI, to Level 5, envisioning systems capable of managing and executing entire organizational operations—covering different tiers of AI capability.
Below are the five AI levels defined by OpenAI:
1. Chatbot: AI with conversational language skills
2. Reasoner: AI with human-level problem-solving ability
3. Agent: Systems capable of taking actions
4. Innovator: AI that assists in invention and creation
5. Organizer: AI capable of performing organizational work
OpenAI believes "Strawberry" is approaching Level 2, which involves problem-solving comparable to a PhD-level researcher without tools. This framework aims to provide a structured approach to understanding and developing AI systems, ultimately surpassing human intelligence.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
