# Claude Opus 5.5 Review: Why It Won Back a Skeptical User

Anthropic's latest AI model delivers speed, cost savings, and surprising new capabilities.

- **Source:** [Lenny's Newsletter](https://www.lennysnewsletter.com)
- **Published:** 2026-09-22
- **Topic:** [Data & AI](https://brifing.si/c/data-ai)
- **Original article:** <https://www.lennysnewsletter.com/p/i-left-claude-for-months-opus-55>
- **On Brifing:** https://brifing.si/s/905a6af6-2bc1-4fe0-9534-87d7be3617fd

## The Claude Exodus

Many power users abandoned Claude due to excessive verbosity, unnecessary disclaimers, and preaching tone in responses. The author switched to OpenAI's Codex for daily tasks, finding it more efficient despite Claude's capabilities. This reflects a growing user preference for concise, task-focused AI interactions over overly cautious language models.

## Opus 5.5 Breakthroughs

Anthropic's Opus 5.5 offers 40% cost reduction over Opus 5 with significantly faster response times. The update introduces a fundamentally different alignment approach that reduces annoying behaviors while maintaining safety. These improvements address core user complaints that drove migration to competitors like OpenAI's offerings.

## Agentic Task Performance

In testing four long-running agentic tasks, Opus 5.5 demonstrated improved reliability, including handling an attempt to manipulate it mid-run. The model showed better context retention and task persistence compared to previous versions. These capabilities make it more suitable for complex, multi-step workflows that require sustained attention.

## Frontend Prototyping Edge

Opus 5.5 has become the author's preferred tool for frontend prototyping, outperforming other models in generating functional code. It handles ChatPRD homepage redesigns effectively while still having some limitations. This suggests specialized capabilities in UI/UX implementation that weren't previously highlighted.

## Unexpected Capability

The model demonstrated one completely unanticipated capability that no other in the author's stack could match. While unspecified, this unique function suggests Anthropic is exploring niche specializations beyond general language tasks. Such differentiation could reshape competitive positioning in the AI market.

## Clear Boundaries

Opus 5.5 showed firm refusal capabilities when pushed beyond its limits, revealing Anthropic's practical safety boundaries. These guardrails appear more thoughtfully implemented than previous versions' preachy disclaimers. The model balances utility with responsible constraints more effectively now.

## Codex Comparison

Despite improvements, Codex still outperforms Opus 5.5 in certain areas, leading the author to maintain a multi-model stack. The testing revealed complementary strengths that suggest continued AI tool diversification. Users gain most value by strategically deploying different models for specific task types.

## Pricing Impact

The 40% cost reduction makes Opus 5.5 particularly compelling for agentic workflows that require many sequential prompts. Lower pricing enables broader experimentation with complex AI-assisted processes. This economic factor may drive more enterprise adoption than pure capability improvements alone.

## Sources covering this story

- https://www.lennysnewsletter.com/p/i-left-claude-for-months-opus-55 — 

---

_Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._