VT-Super 120B: a Vietnamese LLM for business documents

Document AI is only useful if it truly understands the language and business context — and for Vietnamese administrative text, that is not easy. In mid-2026, Viettel AI announced VT-Super 120B, a Vietnamese large language model with 120 billion parameters, trained on administrative, business and specialized document data.

Why this is good news

• A strong Vietnamese LLM that understands administrative text makes document reading and extraction more accurate.
• As a domestic product, it fits the sovereign-AI direction — runnable on in-country infrastructure (see sovereign AI infrastructure).
• It shows Vietnamese business AI no longer depends entirely on foreign models.

The link to enterprise document AI

For sensitive internal documents, what matters is not only a strong model, but where it runs. With a good Vietnamese LLM and in-country infrastructure, running document AI on your own systems becomes practical — in the spirit of the AI Law 134/2025.

Molly Reader

Molly Reader digitizes and extracts document data with local AI — data never leaves your systems. For advice, book a consultation.

Note: model figures are per the vendor's announcement; verify for your specific use.