Just about half a year after OpenAI revolutionized the tech sector with the release of ChatGPT (GPT-3.5) in late November 2022, the company launched GPT-4. However, while OpenAI, led by Sam Altman, continues refining its latest language model, GPT-5 is already in the works.
In this article, we gather all available information about GPT-5, OpenAI’s next major artificial intelligence project, including its release date, functionalities, and key improvements.
When Will GPT-5 Be Released?
There is no official release date for GPT-5. It is also no longer relevant to use the five-month gap between GPT-3.5 and GPT-4 as a reference. However, Sam Altman has provided some hints.
In August 2023, the OpenAI CEO stated that they were not yet training GPT-5 and would not be doing so for some time. In September, he reiterated that “we still have a lot of work to do before we reach GPT-5. It takes a lot of time: we are not close yet. More safety audits are needed.”
However, not training it at that time does not mean they are not doing so now. In fact, GPT-5 was already in the pipeline: in July 2023, OpenAI applied to trademark GPT-5 with the U.S. Patent and Trademark Office. The resolution process could take months.
Recent signs suggest that OpenAI has accelerated its efforts. According to Business Insider, by March 2024, some enterprise clients had already received preliminary versions of GPT-5 and improved ChatGPT versions based on it.
Is GPT-5 Codenamed Orion?
According to The Verge, OpenAI’s next major model will be called Orion and is expected to debut in December 2024. However, in February 2025, it was confirmed that Orion corresponds to GPT-4.5, the company’s last “non-chain-of-thought” model. Despite this, Orion might still be the foundation for GPT-5.
Looking at past releases:
- GPT-3: May 2020
- GPT-3.5: November 2022
- GPT-4: March 2023
Considering this timeline, rumors suggest GPT-5 could be released in the coming months. However, OpenAI has slowed innovation in favor of improving profitability and ensuring safety.
What New Features Will GPT-5 Have?
The GPT-5 trademark application with the U.S. Patent and Trademark Office indicates that OpenAI’s next model will encompass various technological fields, from “downloadable software for language models” to software for “artificial speech and text generation.” Other possible features include:
- Natural language processing, generation, understanding, and analysis
- Text and voice translation between languages
- Machine learning and predictive analytics software
Multimodality
In January 2024, Sam Altman stated on Bill Gates’ podcast Unconfuse Me that OpenAI aims to expand ChatGPT’s capabilities. This includes voice interaction, image and video tools, and a greater emphasis on multimodality, similar to Google’s Gemini Ultra.
Reasoning Capabilities
Altman also emphasized improving reasoning abilities, acknowledging that GPT-4’s reasoning remains limited. He expects GPT-5 to be significantly better in this area.
At the World Government Summit in Dubai, Altman reiterated that GPT-5 would be:
- Smarter
- Capable of multimodal interaction
- Faster in response times
Additionally, there is speculation about the integration of Sora, OpenAI’s video-generation model, into GPT-5.
Reliability and Customization
Another focus for GPT-5 is improving reliability. Altman mentioned that while GPT-4 answers correctly most of the time, occasional errors still occur. GPT-5 aims to significantly reduce such inconsistencies.
Furthermore, OpenAI wants GPT-5 to be more adaptable and customizable to cater to individual and business needs. This includes:
- Personalized interactions
- Integration with personal data like calendars and emails
Codenamed Strawberry
Reuters reports that OpenAI is developing GPT-5 under the codename Strawberry. One of its key advancements is autonomous internet browsing and planning capabilities. Internal documents suggest that Strawberry has exceeded 90% scores on MATH datasets, a benchmark for mathematical reasoning.
In a July 2024 internal OpenAI meeting, Bloomberg reported that Strawberry exhibits human-like reasoning abilities. Additionally, OpenAI has confirmed that o3, initially planned as an independent model, will now be integrated into GPT-5. This means GPT-5 will incorporate:
- Voice interactions
- Canvas-based interfaces
- Deep research tools
How Will GPT-5 Differ from GPT-4?
From the previous points, key differences include:
- Enhanced accuracy and reliability
- Speech input and output capabilities
- Video processing features
- Improved understanding and processing power
According to Altman, “We will be on a steep improvement curve for the next 5 to 10 years; this is the dumbest these models will ever be.”
Increased Computational Power
Enhancing comprehension and reliability will require:
- Larger data volumes
- Greater processing power
- More advanced hardware infrastructure
GPT-3.5 was praised for its intelligence, but GPT-4 went further by leveraging Microsoft’s Azure cloud structure and NVIDIA A100 GPUs for near-human performance.
GPT-5’s architecture remains unknown, but speculation suggests it will surpass human intelligence in specific tasks.
Will GPT-5 Be Free or Paid?
It is too early to determine OpenAI’s official pricing plans, but based on its current strategy:
- ChatGPT-3.5 remains free
- GPT-4 requires a paid subscription (ChatGPT Plus, Professional, or GPT-4 Turbo)
- Microsoft Copilot provides free access to GPT-4
Following this pattern, GPT-5 will likely be a paid model.
Final Thoughts
GPT-5 is shaping up to be a game-changing AI model, with significant advancements in multimodality, reasoning, reliability, and customization. OpenAI is taking a more cautious approach to ensure safety and profitability, but its accelerated progress hints at an imminent release.
While the official launch date remains uncertain, AI enthusiasts and businesses alike can expect a powerful and innovative model that will redefine artificial intelligence applications.