We add complexity to a translation solution only when a simpler solution cannot meet the requirements.

The framework defines eight levels of solution complexity. Fine-tuned models are out of scope in the 2026 edition.

1. Off-the-shelf NMT: A neural machine translation engine with default settings and no glossary.
2. Adjusted NMT: An NMT solution with glossaries, tone of voice, tag handling and other settings supported through the API.
3. Off-the-shelf LLM: An LLM with a simple translation prompt.
4. Adjusted LLM: An LLM with glossary retrieval and prompts tailored to tone of voice, tags and other requirements.
5. Fine-tuned models: NMT domain adaptation or LLM fine-tuning, combined with other customization features. Out of scope in the 2026 edition.
6. Two-step solution: A translation step using NMT or an LLM, followed by automatic post-editing with an LLM to address unmet requirements.
7. Multi-step: More than two steps in one workflow: translation, validation and repair.
8. Multi-agent systems: Agents responsible for different groups of requirements work together to meet all of them.