Google Waxay Soo Fasaxday Gemini 3.8 Flash Waana Speed Waali Ah

Videogan waxaan kaga hadlayaa model-ka cusub ee Gemini 3.8 ee ay Google soo saartay. Maalmihii la soo dhaafay si isdaba joog ah ayay Google models-ka u soo saarayay, kan cusubna wuxuu ku kala baxay laba nooc: Flash oo loogu talagalay coding-ka iyo general use cases-ka, iyo nooca Cyber oo loogu talagalay dhinaca security-ga, gaar ahaan difaaca. Waxa ugu caansan models-ka Gemini waa speed-ka, task-iga aad u dirtidna si quick ah ayuu kuugu dhammaynayaa adigoo aan waqti badan ku luminayn.

Qaybta iigu soo baxday marka aan model-ka fiirinayay waa multidisciplinary expert reasoning. Micnaheedu waa markaad su'aal weydiiso ama task u dirtid inuu tallaabooyinka ka fikiro intaanu jawaab ku siinin, halkii uu isla markiiba wax iska alifi lahaa sida models-kii hore. Benchmark-yada marka la eego wuu ka garaacay models-ka kale, Opus iyo Sonnet-na neck on neck ayuu la socdaa. Qiimuhuna aad buu u raqiis yahay: million token oo input ah waa $0.70, million token oo output ahna waa $3.75. Million in-ka waa su'aasha aad qorto oo aad u dirtid, million out-kana waa waxa model-ku soo saaro. Sababta output-ku uga qaalisan yahay waa computation-ka, reasoning-ka, generation-ka iyo tools-ka uu isticmaalo.

Marka la barbardhigo models-ka sida Fable iyo GPT oo million-ka input-ka $10 ka qaada, output-kana $50, farqigu waa mid weyn. Waxaad ku isticmaali kartaa dhammaan products-ka Google sida Antigravity, Google AI Studio iyo Android Studio. Aniga waxaan u arkaa model promising ah ee aan kuu recommend-garaynayo inaad tijaabiso, gaar ahaan qaybta design-ka aad bay u fiican tahay.

A three-minute Somali-language (Af-Soomaali) AI news video from Dugsiiye on Google's Gemini 3.8 release. It ships in two variants: Flash for coding and general use, and a Cyber variant aimed at security defense. Mohamed walks through what stands out: the speed Gemini models are known for, and multidisciplinary expert reasoning, where the model works through the steps of a task before answering instead of blurting out a reply. On benchmarks it beats or sits neck and neck with the leading models, and the pricing is the headline: $0.70 per million input tokens and $3.75 per million output, against roughly $10 and $50 elsewhere. He explains what million in and million out mean and why output costs more, then covers using it across Antigravity, Google AI Studio and Android Studio, and recommends it especially for design work. Free on YouTube.