Gemini models were announced on December 6th. Gemini Ultra slightly outperforms GPT-4 across the suite of benchmarks, with some caveats (ex. CoT@32 vs. CoT@8).

Was the Gemini Ultra choice and announcement at this state intentional to slow the capabilities race? This questions resolves to "Yes" if evidence surfaces that the Gemini Ultra model was chosen to be announced in order to not hasten the capabilities race. If more capable models were produced but not announced instead, this does not necessarily resolve to "Yes", only if evidence from within Google confirms that other models were not released in order to not hasten the capabilities race.

Edit: Resolution date extended to 12/31/2024!