approximately 36 trillion tokens
Alibaba's Qwen3 models were pretrained on approximately 36 trillion tokens covering 119 languages and dialects, nearly twice the amount used for Qwen2.5.
Dated 29 Apr 2025Source checked 11 Oct 2026Quote found on its page again 16 Oct 2026Basis: stated
In the source's own words
“Qwen3 uses nearly twice that amount, with approximately 36 trillion tokens covering 119 languages and dialects.”
Where it was read
What it is about
Figures it counts in
It is shown on its own, not counted into a figure.
Cited in this guide