Counterpoint research: Llama used as base for 38% of adapted sovereign LLMs
Counterpoint’s index finds a majority of tracked sovereign models are adapted open-source bases and identifies Llama as the most common foreign foundation for those fine-tuned sovereign models.
In this brief: 2 sections 1 min read
Counterpoint evaluated over 170 sovereign LLMs across more than 55 countries for H1 2026.
56% of sovereign LLMs tracked are 'adapted' (fine-tuned versions of open-source foreign bases); 44% are native foundational models.
Llama accounts for 38% of adapted sovereign LLMs; Alibaba’s Qwen and France’s Mistral follow with 20% and 19% respectively.
Widespread use of Llama suggests many countries favor adapting open foreign models over building from scratch.
Counterpoint notes a trend toward a spectrum of autonomy, with more nations gradually pursuing native foundational models.
Llama’s lead across regions highlights Meta’s influence on global model baselines used in sovereign projects.