
GigaChat lost to Claude and Gemini in Russian language on MERA benchmark
Testing GigaChat reveals the harsh truth about Russia’s place in the global artificial intelligence race. Recent tests on the MERA benchmark showed results that force serious reflection. The Russian model, created specifically for working with Russian language, unexpectedly lost to foreign competitors in its own “native element”.
Particularly noteworthy is that the tests were conducted by the GigaChat team itself on a specialized task set developed by the domestic Alliance in the field of artificial intelligence. In comparative tests, the American models Claude 3.7 Sonnet, Gemini 2.0 Flash, Gemini 1.5 Pro. As well as the Chinese DeepSeek-V3 and DeepSeek-V3.0324 demonstrated higher performance.
Jack Clark, co-founder and head of policy at Anthropic, and formerly director of policy at OpenAI, gave an unambiguous assessment of the situation. In his opinion, if GigaChat reflects the state of the entire Russian language model ecosystem, then the country will not only be unable to compete at the cutting edge, but will also face difficulties even in the niche of small open models.
The data from this testing was published in a scientific paper on arxiv. These results indicate that the struggle for leadership in artificial intelligence remains a battle only between the USA and China.