A revised benchmark puts Drex Max v1.1 and Liquid AI’s d1-ultra within three points of each other, while showing they suit different jobs. OrcaRouter’s comparison says the choice comes down less to a headline ranking than to whether a system should make a decision or know when to decline.
Watch Desk analysis
What happened
OrcaRouter compares the two hosted multimodal decision models using Artificial Analysis’s Decision Index v2. It says d1-ultra, previously promoted against an older index, now ranks fifth among 17 models, at 70.45; Drex Max scores 73.39. The report says the earlier eight-point lead claim no longer applies to the revised index, which scores models on new data.
In the comparison, Drex Max scores higher on document decisions and consistency when questions are rephrased. d1-ultra scores higher on the report’s omniscience decision measure and abstains more often when it cannot answer. OrcaRouter also says Drex is faster and less costly per task in the comparison. These are the article’s reported benchmark results and interpretation, not a universal verdict for every workload. Read OrcaRouter’s comparison.
Why it matters
For teams choosing a decision model, a leaderboard score can hide the behaviour that matters in production. A system that declines uncertain questions may be preferable when wrong answers are costly; another may suit document workflows where a usable decision is the point. The comparison argues that teams should test models against their own tasks, rather than treating a small composite-score lead as a settled hierarchy.
Our read
The useful finding is not that one model wins. It is that “better” depends on what failure looks like in your pipeline. OrcaRouter’s comparison makes a case for checking document performance, consistency and abstention separately before committing. A benchmark can help narrow the field; it cannot tell you which mistake your business can afford.
What to watch
- Whether Artificial Analysis publishes further details on the revised index and its component measures.
- How the models perform on the same real-world tasks, especially document workloads.
- Whether pricing and performance change as the services develop.
Discussion spark: When a decision model is unsure, would you rather it make its best call or abstain and pass the work to a person?
Sources and evidence
- Drex Max v1.1 vs Liquid AI d1-ultra: Two Multimodal Scorers, and a Headline That No Longer Holds – OrcaRouter (9 November 2026, 21:28 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.