Claude 3.5 Sonnet Review — Anthropic's Best Model Yet

When Anthropic released Claude 3.5 Sonnet, it was a significant jump — not just an incremental improvement. It outperformed GPT-4o on several benchmark categories and, more importantly, felt noticeably better in real use. Here's what actually changed and what it means for how you use it.

What's New in 3.5 Sonnet

Writing quality — the biggest upgrade

Claude was already the best AI writing assistant for most use cases. 3.5 Sonnet pushed that further. The model handles tone, voice, and nuance better than its predecessor. Long-form outputs (reports, articles, proposals) hold structure more consistently without the drift you see in GPT-4o on extended outputs. It also feels less like "AI output" — the phrasing is more natural and less prone to the patterns that make generated text easy to spot.

Coding — now genuinely competitive

Claude's code generation was always good for explanation and understanding, but 3.5 Sonnet made it a serious contender for actual development work. It holds larger codebases in context better, generates fewer hallucinated function names, and is better at multi-file reasoning. In the Cursor IDE, Claude 3.5 Sonnet is now the default model choice for many developers — not just an option.

Instruction following

This is where Claude consistently outperforms competitors. If you give it complex, multi-part instructions — "do X, but don't do Y, format it as Z, skip the introduction" — 3.5 Sonnet follows all of them. GPT-4o has a tendency to be "helpfully" creative in ways you didn't ask for. Claude stays on task.

How It Stacks Up: Our Ratings

Where It's Still Behind GPT-4o

• No real-time web access on most queries — if you need current information, use Perplexity or ChatGPT with Browse
• Feature surface: ChatGPT has DALL·E image generation, voice mode, more third-party integrations, and a larger plugin ecosystem
• STEM reasoning: OpenAI's o1 and o3 models still hold the edge on complex math and science problems. Claude 3.5 Sonnet is not designed to be a "reasoning" model in that sense

Who Should Switch or Try Claude?

If you're primarily using ChatGPT for writing, document work, or any task where instruction-following and tone quality matter — try Claude 3.5 Sonnet. The $20/mo Pro plan is the same price as ChatGPT Plus. A week of use will tell you whether it fits your workflow better.

If you rely on ChatGPT's image generation (DALL·E), voice features, or specific third-party plugins, those aren't available in Claude — that may be a dealbreaker depending on your workflow.

For most professional writing and knowledge work use cases: Claude 3.5 Sonnet is the best model available right now.