Wire flash
Drex 1.5 decision model ships with 128K token context window, 93.4% accuracy
Editorial responsibility
- No named human review is recorded for this page.
- Source reporting is collected, normalized, translated or condensed automatically when needed.
- Automatically published source-backed update
Drex 1.5, a decision-focused AI model, has been released with a 128,000-token context window, a significant upgrade from its previous limitation of snippets or summaries. According to the announcement, the model can now process entire contracts, full sessions, and larger candidate pools in a single pass. The developer reports 93.4% accuracy across context lengths from 32k to 128k tokens. Drex 1.5 also holds the top position on the Decision Index 0.2.1 benchmark with a score of 58.28. The model is described as a decision engine rather than a chatbot, taking facts and possible choices as input and outputting probabilities for each option without rambling or generating alternatives. The post claims no other model in this category offers a comparable context window.
Source report
Drex 1.5 has just shipped a 128K context window — that's 128,000 tokens on a decision model, and as far as I know, nothing else in this category comes close.
What Is Drex?
If Drex is new to you, think less of a chatbot and more of a decision engine:
- You provide the facts and the possible choices.
- Drex returns a probability for each option.
- No rambling output, no JSON gymnastics, no made-up alternatives.
The Game-Changing Update
The major limitation until now was context. You could give it a snippet, a summary, or maybe part of a document.
That just changed.
Now you can feed it:
- Whole contracts
- Full sessions
- Much larger candidate pools
All in one go.
Performance Metrics
- 93.4% accuracy from 32K through 128K context
- #1 on Decision Index 0.2.1 with a score of 58.28
Decision models are getting serious.
Source
DataChazNeutral / independent