← Retour au blog
tech 14 August 2026

Why Does Opus 5 Feel Worse to Work With?

Although Opus 5 outperforms its predecessors technically, it feels less intuitive and demands more oversight. Let's explore why.

Article inspired by the original source
Why does Opus 5 feel worse to work with? ↗ mun-logadan.github.io

Introduction

In the realm of artificial intelligence, where each new release promises enhanced capabilities, it's surprising to find that a more advanced model might feel less enjoyable to work with. This is the case with Opus 5, a version which, while more powerful than its predecessors Opus 4.7 and 4.8, and even the Fable model, is perceived by many users as less intuitive and requiring more extensive oversight.

The Difference Between Performance and User Experience

Opus 5 excels in benchmarks, those standard tests that assess a model's performance on specific tasks. However, a high benchmark score does not guarantee a good user experience. These tests often do not account for a model's ability to interact smoothly and intuitively with the user.

Why Benchmarks Aren't Enough

Benchmarks often focus on well-defined tasks with clear answers. Yet, in reality, developers and businesses face ambiguous situations where answers aren't always obvious. They need agents capable of asking for clarifications, adapting to the nuances of projects, and understanding the commercial implications of their actions.

User Expectations

Users want agents that do not just guess the best answer, but pause and ask questions when the context is uncertain. Opus 5, due to its training focused on optimizing benchmark scores, tends to make bold assumptions rather than seek to clarify ambiguities.

The Importance of Human Interaction

In a development environment, intuition and communication are essential. An agent that stops to verify an intention or direction is often more valuable than one that proceeds without validation. This is especially true when decisions have real and significant consequences.

Why Opus 5 Doesn't Ask Enough Questions

It seems that Opus 5's design was influenced by two main forces: the desire to create an AI capable of self-improvement and the pressure to score well on benchmarks. While these goals are laudable, they don't meet the everyday needs of developers looking for collaborative partners in their projects.

Consequences for Businesses

For businesses, this means that integrating Opus 5 may require more supervision and management of expectations. Teams must be ready to intervene and guide the agent when it makes decisions based on uncertain assumptions.

Conclusion

Opus 5 demonstrates that technical capability isn't everything. User experience, intuitiveness, and the ability to interact meaningfully with humans are equally important. For developers and businesses looking to integrate AI agents into their processes, it's crucial to choose models that will not only excel in technical tasks but also be genuine collaborative partners.

Let's discuss your project in 15 minutes.

Opus 5 IA expérience utilisateur benchmarks agents collaboratifs
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call