OpenAI presented the most powerful GPT 5.4 mini and nano models
OpenAI company has presented the GPT-5.4 mini and GPT-5.4 nano models. This was reported by Zamin.uz.
They were evaluated as the most powerful small artificial intelligence models created so far. The new models have architectural features designed for speed, efficiency, and heavy workloads.
The GPT-5.4 mini model stands out with its capabilities in programming, logical reasoning, multimodal analysis, and the use of various tools. The company announced that this model is significantly superior to the GPT-5 mini version, working more than twice as fast.
In benchmark tests, the GPT-5.4 mini showed results of 53.40% in the SWE-bench Pro test, 59.30% in Terminal-Bench 2.0, and 70.60% in the OSWorld-Verified test. These indicators show that it is close to the full GPT-5.4 model.
The company emphasized that the mini model was developed for low-latency AI assistants, lightweight subagents, interface control systems analyzing screenshots, and real-time multimodal applications. According to OpenAI, in such cases, the largest model is not always the best solution.
The main focus is immediately on responding and working reliably with tools. The GPT-5.4 nano was presented as the smallest and fastest model in this series.
It is designed for tasks where speed and cost are important, including classification, data extraction, ranking, and simple assistant workflows. OpenAI also announced initial benchmark results for the nano model.
Among them, 52.39% in the SWE-bench Pro test and 82.83% in the GPQA Diamond test were recorded. The company also reported that its clients are testing both models in production environments.
