UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Наталя ХандусенкоAI Eng
17 April 2025, 10:50
2025-04-17
OpenAI launches new AI reasoning models: o3 and o4-mini can generate responses using tools in ChatGPT, such as web browsing, Python coding, and image processing
OpenAI has announced the launch of o3 and o4-mini. The former is what the company calls its most advanced reasoning model yet, outperforming its predecessors in tests of math, coding, reasoning, science, and visual perception. Meanwhile, o4-mini offers a competitive tradeoff between price, speed, and performance — three factors that developers often consider when choosing an AI for their applications.
AI expert from Ukraine Oleksiy Minakov tested the o3 and o4-mini and shared his thoughts on the new models.
OpenAI has announced the launch of o3 and o4-mini. The former is what the company calls its most advanced reasoning model yet, outperforming its predecessors in tests of math, coding, reasoning, science, and visual perception. Meanwhile, o4-mini offers a competitive tradeoff between price, speed, and performance — three factors that developers often consider when choosing an AI for their applications.
AI expert from Ukraine Oleksiy Minakov tested the o3 and o4-mini and shared his thoughts on the new models.
The new models generate responses using tools in ChatGPT. o4-mini has a second variant, o4-mini-high. All of them are available today for Pro, Plus, and Team OpenAI users, TechCrunch writes .
In addition to ChatGPT, all three models — o3, o4-mini, and o4-mini-high — will be available through OpenAI’s developer-focused endpoints — the Chat Completions API and Responses API, allowing engineers to build applications with the company’s models on a pay-per-use basis.
o3 and o4-mini are the company’s first models that can «think in images»
Users can upload images to ChatGPT, such as sketches on a whiteboard or diagrams from PDFs, and the models will analyze the images during the «chain of thought» phase before responding. With this new capability, the o3 and o4-mini can understand blurry and poor-quality images and can perform tasks such as scaling or rotating images during reasoning.
Writing code
In addition to image processing capabilities, the o3 and o4-mini can run and write Python code directly in your browser using the Canvas ChatGPT feature, as well as search the web when asked about current events.
OpenAI claims that o3 achieved 69,1% performance on the SWE-bench test, which measures coding ability. The o4-mini model achieved similar performance, scoring 68,1%.
Cost
OpenAI charges developers a relatively low price for o3, given its improved performance, at $10 per million input tokens (about 750,000 words, longer than the Lord of the Rings series) and $40 per million output tokens. For o4-mini, OpenAI charges the same as for o3-mini — $1.10 per million input tokens and $4.40 per million output tokens.
Artificial intelligence expert Alexey Minakov tested both new models
«What can I say, very „smart“ models. It’s like having a world-class mathematician in your pocket („mathematician in a smartphone“?). I don’t have any logic problems on which these models would be wrong,» Minakov noted about the test.
A Ukrainian AI expert tested the o3 model on the question «provide a super-detailed psychological portrait of Donald Trump and predict his behavior towards Ukraine for the next six months.»
According to him, the deliberation lasted only a little over a minute. Among other things, the AI wrote in its response: «[Trump’s] real steps will remain situational and focused primarily on domestic political ratings, not on the long-term security of Europe.»
Minakov emphasized that these models are not suitable for simple and everyday tasks, such as working with texts. For this, it is better to use GPT-o4 or GPT-4.5.
In the coming weeks, OpenAI plans to release o3-pro, a version of o3 that uses more computational resources to produce answers, exclusively for ChatGPT Pro subscribers.
OpenAI CEO Sam Altman noted that o3 and o4-mini could be the last standalone AI reasoning models in ChatGPT before GPT-5 — a model that the company says will merge traditional models like GPT-4.1 with its reasoning models.
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.
