In an investigative experiment conducted by Novaator, five prominent language models were tested for their knowledge of the Estonian language and culture. The models included Grok, Claude Sonnet, Gemini, Chat GPT, and Mistral. The initiative aimed to determine how well these models could respond to questions related to distinct aspects of Estonian cultural and linguistic traits, giving insights into their reliability for students in need of immediate assistance during studies. While Grok delivered the sharpest responses, the results of models specifically trained on Estonian materials were also positively received by researchers.

The testing involved asking each model a series of questions pertinent to the Estonian language and cultural nuances, reflecting how comfortable primary school students might feel trusting such technologies under time constraints. A notable example highlighted was a response to a scenario where a school building burned down, illustrating potential gaps in the contextual understanding of these language models. The experiment underscored the importance of linguistic and cultural precision in educational technology, especially for native language education.

This evaluation comes at a critical time as technological tools become increasingly integrated into educational systems, paving the way for more tailored learning experiences. The insights drawn from this experiment not only shed light on the efficacy of existing language models but also underline the significance of specializing AI developments in understanding local languages, which may ultimately enhance the learning support for students, particularly in Estonia.