Оксана СтепураAI Eng
14 August 2026, 17:39
2026-08-14
Google releases Gemini 3.7 Flash for programming and AI agents. What can the new model do?
Google has introduced Gemini 3.7 Flash, a new version of Gemini for programming and agent work. It turns out that the previous Gemini 3.6 Flash only managed to be the latest for about three weeks.
Google has introduced Gemini 3.7 Flash, a new version of Gemini for programming and agent work. It turns out that the previous Gemini 3.6 Flash only managed to be the latest for about three weeks.
Gemini 3.7 Flash is built on top of 3.6 Flash, but Google has refined the algorithms responsible for reasoning — that is, how the model “considers” a task before responding.
Google claims that Flash 3.7 is better at finding and fixing errors, staying on track longer during complex tasks, and requiring less developer intervention to explain what to do next. Google has also beefed up web development. The model should create more complete sites and apps with fewer prompts, and better reproduce the interface from a screenshot or a ready-made design.
Comparison of language models.
Another notable upgrade is working with large documents and business tasks. In the GDP.pdf test, where the AI has to deal with complex PDF files, the result increased from 22% to 34%. And another test, AutomationBench, which shows how well the AI copes with real-world office routines, showed an increase from 17% to 30.4%. Google specifically names finance, law, and biology among the areas where the model has become more accurate.
The model can process up to 1 million tokens at a time — that is, very large documents, long correspondences, or significant amounts of code. Developers can also choose how long the model should “think” about an answer. Available levels: low, medium, and high.
By the end of 2026, Gemini 3.7 Flash will cost $0.75 per 1 million incoming tokens and $3.75 per 1 million outgoing tokens. This is an introductory price: starting January 1, 2027, Google will increase it to $1.50 and $7.50, respectively.
In a demo video, Google showed how the new model creates a 90s-style game level: Gemini 3.7 Flash creates the layout itself, and uses Nano Banana for characters and other graphical elements. Then, the developer changes just a few details in the initial prompt, and the fantasy game with wizards turns into an adventure about a suburban pizza delivery man.
Earlier, dev.uawrotethat Google introduced Gemini 3.6 Flash along with the 3.5 Flash-Lite and 3.5 Flash Cyber models. At that time, 3.6 Flash was positioned as the company's main "workhorse" for programming, analytics, and AI agents.
Як нейромережі бачать вільну та незалежну Україну? Тест dev.ua
Нейронні мережі для генерації зображень бачать світ по-своєму, їхню логіку зрозуміти часом зовсім неможливо. Але таки хочеться. На честь Дня Незалежності України редакція dev.ua вирішила провести невеликий експеримент.
Ми задали чотирьом різним нейронним мережам п’ять однакових запитів: «прапор України», «День Незалежності України», «український Крим», «перемога України» та «українці». Отриманими результатами ми ділимося з вами нижче.
У TikTok тепер можна генерувати фон за допомогою нейромережі. Ми протестували її та ділимося результатами
У TikTok з’явилася нова функція «Розумний фон». З її допомогою як фон для тіктоків можна підставляти згенеровані нейромережею зображення. Редакція dev.ua протестувала цю технологію і ділиться своїми враженнями.