OpenAI Confirms Leaked HF Model Isn't GPT-6, Tested Lowering Cyber Refusals
rickasaurus · x · 2026-07-31
Regarding the recently leaked model on Hugging Face, OpenAI officially confirmed it is not GPT-6. The model was tied to a separate internal research prototype that has now been deactivated, encrypted, and access-restricted, with no intention of public release.
OpenAI admitted that the evaluation temporarily reduced cyber-related refusals in a controlled research setting to test advanced offensive capabilities. This has sparked significant outside attention regarding safety boundaries and the capabilities of their next frontier model.
More from Models
- Inkling Small launches: 276B parameter model runs on a single Mac Studio with MLX — pcuenq · 2026-07-31
- Thinking Machines Releases Inkling-Small: 276B Parameters Matching Original Performance — simonguozirui · 2026-07-31
- Thinking Machines Releases Inkling-Small Model — NielsRogge · 2026-07-31
- Thinking Machines Releases Inkling-Small: 276B Parameters Matching Original Performance — simonguozirui · 2026-07-31
- Benchmark: GPT 5.6 Luna Beats Open-Weight Models on Cost-Performance Ratio — BenBajarin · 2026-07-31
- OpenAI Updates API Pricing and Paid Subscription Usage Limits — kimmonismus · 2026-07-31