UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Наталя ХандусенкоAI Eng
17 December 2025, 15:15
2025-12-17
Xiaomi launches new open-source AI model MiMo-V2-Flash to compete with DeepSeek and OpenAI
Chinese smartphone and electric car giant Xiaomi has released a new artificial intelligence model, MiMo-V2-Flash. The company claims that it demonstrates the best results in solving complex logic problems, writing code and working in autonomous agent mode, but is also ideal for the role of an assistant for everyday tasks.
Chinese smartphone and electric car giant Xiaomi has released a new artificial intelligence model, MiMo-V2-Flash. The company claims that it demonstrates the best results in solving complex logic problems, writing code and working in autonomous agent mode, but is also ideal for the role of an assistant for everyday tasks.
MiMo-V2-Flash is available worldwide on the Xiaomi MiMo Studio developer site, the Hugging Face platform, and via APIs, writes the South China Morning Post.
Xiaomi claims that the MiMo-V2-Flash model is designed with a focus on performance.
“It provides lightning-fast withdrawal speeds of 150 tokens per second, while maintaining an ultra-low cost of $0.1 per million incoming tokens and $0.3 per million outgoing ones. This makes it one of the most cost-effective high-performance models on the market,” the company noted.
According to Xiaomi, their development is not inferior to such giants of reasoning as Kimi K2 and DeepSeek V3.2 in logic tests, but at the same time copes better with creative writing. A separate advantage is called working with large amounts of data: in the evaluation of long contexts, MiMo-V2-Flash bypassed the model from Moonshot AI.
For agent tasks, Xiaomi claims that MiMo-V2-Flash scored 73.4% on SWE-Bench Verified, outperforming all open-source competitors, coming close to the performance of OpenAI’s GPT-5-High. It was also on par with Claude 4.5 Sonnet, which is considered the best coding model in the industry, at a price that is significantly lower than the cost of the Anthropic model.
MiMo-V2-Flash was developed using the Mixture-of-Experts (MoE) architecture, an approach in machine learning that divides a model into separate subnetworks, or “experts,” each specializing in a specific part of the input data to collectively perform a task. The total number of model parameters is 309 billion.
Recall that Xiaomi launched its first AI model, MiMo, in May of this year.
There's nothing to breathe. How to determine the level of air pollution and what to do when the city is drowning in smoke after shelling
Kyiv residents are once again experiencing air quality problems in the Ukrainian capital: visibility has significantly deteriorated, a strong smell of smoke has appeared in certain areas, and breathing has become difficult. The reason for this phenomenon is another Russian shelling, which has caused numerous fires in various areas of the capital and in some places has led to pollution levels exceeding the norm by dozens of times. We remind you of what you need to know about air quality and what useful tools are available for monitoring it.
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.