Sovereign AI Isn’t so Sovereign: Meta’s Llama Powers 38% of Adapted Models

Meta Platforms Inc. has been identified as the leading foundation for sovereign artificial intelligence models, with its Llama model accounting for 38% of adapted sovereign large language models (LLMs) tracked by Counterpoint Research. The report indicates that 56% of sovereign LLMs are adapted models, while the remaining 44% are foundational models trained on native data. Competitors include Alibaba's Qwen with a 20% share and France's Mistral with 19%. The research evaluated over 170 models across more than 55 countries, highlighting a trend towards greater AI autonomy and diversification in AI technology stacks.