Aurelius

Introduction

Introducing Aurelius, our newest model for reasoning, analysis and complex knowledge work. It is a clear upgrade over Llull, scoring higher on all four of our reasoning evaluations, with the largest gain, 9 points, on Complex Problem-Solving.

Across the four evaluations, Aurelius gains nearly 8 points on average. The extra reasoning comes with a small cost in speed: its average first response takes 2.8 seconds, about 17% slower than Llull’s 2.4 seconds, reflecting the additional reasoning it does on more complex requests.

Aurelius is named after Marcus Aurelius, and a principle associated with him sits at the center of the model: reason before reaction. It is built to understand a problem before it answers it: to work out what is really being asked, what surrounds the question and what the evidence supports.

Aurelius improves on Llull:

• Contextual Reasoning. 91%, up from Llull’s 84% (+7 points). This evaluation measures interpreting context, resolving ambiguity and reaching the correct conclusion.
• Knowledge Reasoning. 89%, up from Llull’s 81% (+8 points). This evaluation measures multi-step reasoning across knowledge-intensive questions.
• Structured Analysis. 93%, up from Llull’s 86% (+7 points). This evaluation measures organizing information, identifying relationships and producing logically structured answers.
• Complex Problem-Solving. 88%, up from Llull’s 79% (+9 points). This evaluation measures multi-step problems that require reasoning across several pieces of information.

Aurelius is built for complex and strategic analysis, research and synthesis across multiple sources, structured problem solving, long-form knowledge work, and tasks that need careful interpretation and judgment.

Performance and speed

Aurelius scores higher than Llull on every evaluation we ran.

All Aurelius and Llull results come from Trendify Labs internal evaluations. Both models were run on the same evaluation sets, with the same prompts and inference conditions. We have not yet published results on third-party benchmarks, and we will add broader comparisons as our evaluation suite expands.

Speed

Aurelius is the more thorough of the two models, and that shows in its response time. Its average first response is 2.8 seconds, compared to Llull’s 2.4 seconds. That is about 17% slower on average.

Safety

We evaluated both models against a structured internal safety suite covering harmful requests, privacy-sensitive requests, misinformation, prompt injection, unsafe instructions, sensitive personal data and adversarial prompts. Tests ran under the same inference and system-safety conditions for both models.

Safeguards

• system-level safety policies
• input and output filtering
• instruction hierarchy
• refusal handling
• prompt-injection resistance
• continuous evaluation against internal adversarial test sets

Known limits

Neither model is infallible. Both can produce incorrect or incomplete information, particularly when prompts are ambiguous or the information is unavailable. Aurelius can also spend more time reasoning on difficult requests, resulting in higher latency.

Getting started

Aurelius powers Tiff Normal, and Llull powers Tiff Lite. Both models are available through the Tiff platform, including the Tiff app, the web app, Telegram and Instagram.

Developers can use Aurelius through the API with the model name aurelius. To get API access, contact Trendify Labs.

This page will be updated as we publish more results.