On June 10, Mistral - a French artificial intelligence (AI) startup - introduced a new model called Magistral that is capable of reasoning and solving complex problems.
This move continues the innovation process of leading developers in the field of AI in the US.
According to Mistral's announcement, Magistral is now deployed on the company's platform as well as on Hugging Face - a popular open-source AI platform.
Mistral claims the model is designed for systematic thinking, similar to the way humans think.
The French startup also emphasized that Magistral was developed for general use, requiring longer inference processes and higher accuracy than Mistral's previous generation of large language models (LLMs).
Similar to other reasoning models, Magistral allows users to follow the process of approaching and solving a problem presented in natural language by representing “thought sequences.”
This means that users in fields such as legal, finance, healthcare and governance will receive answers that can be traced back to the source of the argument, ensuring that “every conclusion can be explained in a clear, logical, step-by-step sequence.”
Mistral's statement also highlights one of the top challenges in the AI industry: “interpretability,” or how users can understand the sequence in which an AI system responds to a given problem.
Unlike traditional software that is programmed directly by humans, modern AI models are trained on huge amounts of data. Therefore, how AI systems operate is often a mystery even to technology developers.
Mistral also highlighted Magistral's superior ability in software coding and content creation, areas that require flexible reasoning and language skills.
Other competing inference models include OpenAI's o3, Google's Gemini, Anthropic's Claude, and China's DeepSeek's R1.
While many companies are pushing claims about AI's reasoning capabilities, Apple - a tech giant that is often seen as lagging behind in the AI race - has voiced its doubts.
In a paper titled “The Mind Illusion,” a team of Apple researchers argues that “current models still have fundamental limitations, failing to develop general reasoning capabilities beyond a certain threshold of complexity.”/.
(Vietnam News Agency/Vietnam+)
Source: https://www.vietnamplus.vn/mistral-ra-mat-mo-hinh-ai-co-kha-nang-suy-luan-linh-hoat-post1043652.vnp
Comment (0)