NEWS · MODELS · #1981
OpenAI may ship one or two models this year while flagship still needs alignment work
Employees told The Information that OpenAI expects to release one or two new models before year-end, but its next flagship model remains delayed by alignment and testing work. The model family uses looped transformers (branded internally as Recurrent Depth/Astra), which improve math and coding performance but make chain-of-thought monitoring harder; a Microsoft write-up briefly leaked that GPT-6.1 Sol uses looping inference.
KEY POINTS
- Employees told The Information that OpenAI expects to release one or two new models before year-end, but its next flagship model remains delayed by alignment and testing work.
- The model family uses looped transformers (branded internally as Recurrent Depth/Astra), which improve math and coding performance but make chain-of-thought monitoring harder; a Microsoft write-up briefly leaked that GPT-6.1 Sol uses looping inference.
- This matters because looped-transformer designs can change model capabilities and make safety monitoring and regulatory prelaunch evaluations harder, affecting release timing and oversight.
WHY IT MATTERS
This matters because looped-transformer designs can change model capabilities and make safety monitoring and regulatory prelaunch evaluations harder, affecting release timing and oversight.