Drex v1.5

Drex v1.5 is a decision model from Nace AI. It does not write free text. It reads a state given as text or JSON together with typed questions, and returns a probability for each option in a single pass, so software can act on a bounded answer. It has about 9 billion parameters (8,953,803,264 in BF16) and is finetuned from XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B. The context length is 16,384 tokens by default and can be extended to 131,072. The weights are on Hugging Face, where the repository was created on 28 September 2026, and the code and runners are on GitHub. The licence is Nace.AI Open RAIL-M, a modified Open RAIL-M under which commercial use above set thresholds needs a separate licence, so it is not open source in the OSI sense and a buyer should read it. The hosted API price is not public. The model card reports a Decision Index 0.3.1 score of 58.08 from Nace AI's own run, within a tie band with Jev at 57.96 and other models. These are the vendor's own figures and are not stored as scores. The vendor's blog of 24 September 2026 introduced the first Drex and was updated on 28 September.

Not yet independently verified: The release date of 28 September 2026 rests only on the creation of the Hugging Face repository; the vendor's own announcement date for v1.5 was not found, and its blog post of 24 September 2026 introduced the first Drex. The benchmark figures are Nace AI's own, so no score is stored. The licence is a modified Open RAIL-M, not an open source licence approved by the OSI. No independent report was found. We will update this when it can be confirmed, and remove this note.