Drex
A System One decision model released by Nace, built on the Qwen3.6-35B-A3B MoE architecture. It supports structured probability output and is released as open weights under the Open RAIL-M license, with both hosted and self-hosted deployment options.
Timeline (last 90 days)
Language capability: Jev is mainly English-focused, with lower accuracy in other languages including CJK; Drex has not published a language list; Laya covers more than 100 languages, with 45 of 51 tested languages usable, and has a built-in language router.
Both Jev and Drex support up to 255 options; Jev reaches 79.2% on the full Banking77 test (77 labels); Laya's performance degrades beyond about 20 options under default settings, reaching only 0.425 on Banking77.
On calibration, Jev's ECE in independent tests varies between 0.02 and 0.35 depending on the task; Laya's ECE is 0.466 before temperature fitting (overconfident) and drops to 0.081 after fitting; Drex has not published calibration data.
The three System One decision models share the same API: POST /v1/systemone, passing in state and a question mapping; this protocol is defined by TypeSafe, Jev is the first implementation, and Drex and Laya are wire-compatible.
Drex has open weights but uses the Open RAIL-M license with usage restrictions, so it is not open source as defined by OSI; it offers both hosted and self-hosted channels; it is based on Qwen3.6-35B-A3B MoE, with 35B total parameters and about 3B activated; self-hosted default context is 16K with a maximum of 128K, and hosted is 128K.
In zero-shot quality, Jev is the reference benchmark; Drex's vendor reports that on its Decision Index it leads Jev 61.25 to 57.96, but on JevBench it is slightly lower at 85.7% versus Jev's 86.6%; Laya's zero-shot on typed-decisions is only 0.362, below the 0.461 majority-class baseline.
After fine-tuning, Laya reaches 0.766 on the same benchmark, versus the third-party-published Jev figure of 0.727; Drex offers fine-tuning as a vendor service.
Nace.AI launched Drex on September 24, 2026, claiming it topped the public decision index.
Nace.AI open-sourced a decision model called Drex 1.5
All Facts (8)
Language capability: Jev is mainly English-focused, with lower accuracy in other languages including CJK; Drex has not published a language list; Laya covers more than 100 languages, with 45 of 51 tested languages usable, and has a built-in language router.
50%UnverifiedThe three System One decision models share the same API: POST /v1/systemone, passing in state and a question mapping; this protocol is defined by TypeSafe, Jev is the first implementation, and Drex and Laya are wire-compatible.
50%UnverifiedDrex has open weights but uses the Open RAIL-M license with usage restrictions, so it is not open source as defined by OSI; it offers both hosted and self-hosted channels; it is based on Qwen3.6-35B-A3B MoE, with 35B total parameters and about 3B activated; self-hosted default context is 16K with a maximum of 128K, and hosted is 128K.
50%UnverifiedIn zero-shot quality, Jev is the reference benchmark; Drex's vendor reports that on its Decision Index it leads Jev 61.25 to 57.96, but on JevBench it is slightly lower at 85.7% versus Jev's 86.6%; Laya's zero-shot on typed-decisions is only 0.362, below the 0.461 majority-class baseline.
50%UnverifiedAfter fine-tuning, Laya reaches 0.766 on the same benchmark, versus the third-party-published Jev figure of 0.727; Drex offers fine-tuning as a vendor service.
50%UnverifiedBoth Jev and Drex support up to 255 options; Jev reaches 79.2% on the full Banking77 test (77 labels); Laya's performance degrades beyond about 20 options under default settings, reaching only 0.425 on Banking77.
50%UnverifiedOn calibration, Jev's ECE in independent tests varies between 0.02 and 0.35 depending on the task; Laya's ECE is 0.466 before temperature fitting (overconfident) and drops to 0.081 after fitting; Drex has not published calibration data.
50%UnverifiedNace.AI open-sourced a decision model called Drex 1.5
50%