Liquid AI's d1 tops Jev on Decision Index

Jev arrived in September 2026 and immediately drew wide attention, prompting attempts around the world to link it and similar models with large language models (AI that understands and generates text) and existing apps to speed up task processing. That interest has also set off an active push to build decision models on Jev's ideas, and d1 is one of those efforts.

Liquid AI presents d1 as the first model to outperform Jev on Hugging Face's Decision Index, backed by its own run of the index version 0.2.1, in which d1 scored higher than Jev on many tests. The company's post also points to advantages in multilingual evaluations, robustness against prompt injection, and handling longer inputs, positioning d1 for fast, structured decision-making.

What the release hinges on is whether those results hold up outside Liquid AI's own testing, since a benchmark run carried out by the model's developer is the kind of evidence independent evaluation tends to test. For teams currently pairing Jev-style models with existing applications, d1 appears to offer another candidate to weigh, with the size of its real edge likely to depend on how it performs under third-party tests.