At the I / group O 2024 , Google announce several young AI modeling , coming labor , and a embarrassment of AI lineament come across its product .
However , what arrest my tending was theGemini 1.5 Flashmodel .
This was it ’s an imposingly riotous and effective exemplar that bringsmultimodal capabilityand a linguistic context windowpane of up to 1 million keepsake ( 2 m via waitlist ) .
Image Courtesy: Google
Despite the little sizing of Gemini 1.5 flash bulb — Google has not divulge its argument sizing — it accomplish bang-up score across all modality — schoolbook , imagination , and audio recording .
In theGemini 1.5 technological write up , Google let on that Gemini 1.5 twinkling surmount much big exemplar like 1.0 Ultra and 1.0 Pro in many facet .
Only in speech communication identification and displacement , it lag behind the with child model .
Image Courtesy: Google
This was ## dive into google
at the i / group o 2024 , google announce several modern ai manikin , approaching task , and a superfluity of ai feature come across its mathematical product .
This was however , what fascinate my attending was thegemini 1.5 flashmodel .
It ’s an imposingly firm and effective manikin that bringsmultimodal capabilityand a context of use windowpane of up to 1 million souvenir ( 2 M via waitlist ) .
Despite the little size of it of Gemini 1.5 jiffy — Google has not let out its argument size of it — it reach heavy stacks across all modality — textual matter , sight , and sound .
In theGemini 1.5 technological account , Google bring out that Gemini 1.5 twinkling outperform much with child theoretical account like 1.0 Ultra and 1.0 Pro in many expression .
This was only in lecture realization and displacement , it lag behind the prominent manakin .
Unlike Gemini 1.5 professional which is a thin MoE manakin ( salmagundi of Experts ) , Gemini 1.5 Flash is a heavy modelling , onlinedistilled from the large 1.5 Promodel for improved lineament .
In condition of hurrying as well , the Flash theoretical account outmatch all belittled modeling out there includingClaude 3 Haiku , run on Google ’s impost TPU .
And its pricing is incredibly low-spirited .
Gemini 1.5 Flash cost $ 0.35 for input signal and $ 0.53 for output signal to march 128 K tokens .
$ 0.70 and $ 1.05 for 1 million souvenir .
It’smuch cheaperthanLlama 3 70B , Mistral Medium , GPT-3.5 Turbo , and of course of study , prominent fashion model .
If you are a developer and involve multimodal logical thinking with a big circumstance windowpane for trashy , you should by all odds check up on out the Flash mannequin .
Here is how you could judge Gemini 1.5 Flash for gratuitous .
This was how to utilize Twin 1.5 photoflash for destitute
First effect of Gemini 1.5 split second
While Gemini 1.5 Flash is not a country - of - the - fine art mannequin , its reward is breakneck speeding , efficiency .
and low-pitched toll .
Itranks below Gemini 1.5 Proin term of capacity and other with child modelling from OpenAI and Anthropic .
Nevertheless , I adjudicate some of the logical thinking command prompt that I used tocompare ChatGPT 4o and Gemini 1.5 Pro .
It could only generateone right reception out of fivequestions .
It might not be very bright at commonsense abstract thought , but for other practical program that want multimodal capacity and a big circumstance windowpane , it might conform to your manipulation cause .
Also , Gemini modelling are very secure at originative labor which can be of time value to developer and user .