Google is actively working on an OpenAI o1-style reasoning model named 'Centaur'
On the lmarena, the model consistently answers the hardest benchmark questions correctly.
It does so without showing any work or reasoning but takes about 30 seconds to stream the first token.
On the lmarena, the model consistently answers the hardest benchmark questions correctly.
It does so without showing any work or reasoning but takes about 30 seconds to stream the first token.