OpenAI releases the full GPT-2 1.5B weights
OpenAI releases code and model weights for the largest GPT-2 configuration, concluding its nine-month staged-release experiment.
What happened
Event details
OpenAI released the largest 1.5B-parameter GPT-2 model together with code and model weights. The team had opened only a smaller model in February, followed by 355M and 774M versions in May and August; this release completed the nine-month staged process. The accompanying review found that GPT-2 outputs could be convincing, that fine-tuning could enable misuse and that content-only detection was insufficient on its own. OpenAI also reported no strong evidence of real-world misuse at the time. The full release therefore reflected an evidence-informed policy adjustment, not a conclusion that the risks had disappeared.
Assessment
Why it matters
The release completed an early staged-release experiment and left a documented case for open weights, misuse evaluation, bias analysis and detection limits.
Availability
Access notes
GPT-2 1.5B code and model weights were released publicly.