Despite previous statements that there were no immediate plans for training GPT-5, reports suggest that training was completed in 2023 and the model is expected to launch in 2024. The new model will likely be directed towards OpenAI’s enterprise customers. There have been reports of issues with the current GPT-4 model, including hallucinations and query failures, which some speculate may be due to OpenAI training new, unreleased models alongside the current ones.
Key takeaways:
- OpenAI's next large language model, GPT-5, is expected to be launched within months, with demos currently being sent to enterprise customers for testing.
- The new model may include an AI agent that can execute tasks without human assistance, and it is currently undergoing safety testing, including red teaming.
- Despite rumors of an incremental GPT-4.5, reports suggest that training for GPT-5 had already been completed in 2023 and it would be launched sometime in 2024.
- OpenAI's GPT-5 will likely be directed toward enterprise customers, potentially establishing a tier system similar to Google Gemini LLM tiers, with different model versions serving different purposes and customers.