Google recently announced the availability of a new experimental model for testing on AI Studio. Named Gemini Experimental 1114, the title reflects its announcement date, November 14th. This model is accessible in the preview section via the model selector.
Notably, it comes with a 32k context window, which is significantly smaller than what other Gemini models offer. Additionally, it lacks search grounding capabilities. Early user feedback suggested the model would excel in reasoning tasks; however, it sometimes takes longer to process problems compared to its counterparts.
Massive News from Chatbot Arena🔥@GoogleDeepMind's latest Gemini (Exp 1114), tested with 6K+ community votes over the past week, now ranks joint #1 overall with an impressive 40+ score leap — matching 4o-latest in and surpassing o1-preview! It also claims #1 on Vision… https://t.co/AgfOk9WHNZ pic.twitter.com/HPmcWE6zzI
— lmarena.ai (formerly lmsys.org) (@lmarena_ai) November 14, 2024
Reports from Arena indicate that this model has been under testing for some time. It has now risen to the top of leaderboards, even surpassing OpenAI’s O1 Preview. Previously, it was available under the Gemini test label in Arena’s battle mode.
From my own tests, I didn’t notice significant differences, though this might be due to my high expectations. The model delivers decent outputs but occasionally fabricates answers—an expected limitation, given that it’s likely a smaller model. Notably, it hasn’t been branded as Gemini 2, suggesting it could either be a smaller version of Gemini 2 or a completely new type of model distinct from the core series.
You can explore this model for free on Google AI Studio and form your own impressions.
