Hey there,

2025 has been a constant back-and-forth battle between AI developers and now Anthropic has made its move. Their newest model Claude Sonnet 4.5 has made improvements across the board and is performing well compared to other state-of-the-art models. This week I wanted to take a look at Anthropic's response to GPT-5 and break down what’s changed and how it compares to other models. 

What's New in Claude Sonnet 4.5 ?

Anthropic has created Sonnet 4.5 to be their new go-to model, and after digging in, it's clear why. While this model has seen significant improvements across the board, I want to break down the three areas with the biggest impact: its new agentic abilities, code generation, and content creation. 

Agentic Capabilities: 

Anthropic has made agentic use cases a priority with Sonnet 4.5. Here are a couple of the most impactful changes.

• Task Management: Sonnet 4.5 will now break down a complex issue into smaller, more focused tasks instead of trying to solve the whole problem at once. This is paired with fact-based reporting to help you accurately track status updates.  

• Improved Tool Calls: Sonnet 4.5 can now use multiple tools in parallel, like reading several files at once, allowing it to build a complete picture of a topic much faster.

Coding: 

Code generation is where Sonnet 4.5 really shines. Developers and users alike will benefit from Sonnet’s improved problem solving and code generation. The improvements don’t stop with better code, Sonnet 4.5 is also better at planning the architecture of the app before writing it. It’s also better at identifying vulnerabilities and implementing more robust security practices. Best of all though is that Sonnet 4.5 is way better at sticking to the specific coding instructions you give it. 

Content and Communication: 

One of the worst experiences of working with AI is when they give you long-winded responses without saying much. Thankfully this has been addressed with Sonnet 4.5 by making the model more “concise, direct, and natural”. This is especially noticeable when the model produces vastly better initial responses. 

How to Access Sonnet 4.5

Sonnet 4.5 is free for all Claude users and is the new default model. It is also available within the API. 

How Does It Compare?

The most important question is, how does Sonnet 4.5 compare to other flagship models? For this we’ll look in two places; the benchmark evals and LMArena. As a reminder, LMArena allows users to blindly rank answers from various AI models and is used to gauge the user response.

On LMArena, Sonnet 4.5 is already a top performer, currently ranking #1 in multiple categories, including Text. Being the new default model in Claude means that Sonnet needs to be excellent at understanding and executing text based communication. This model also placed well on the WebDev leaderboard where it tied with models like Gemini 2.5 Pro. 

While it's important to understand how people are rating its responses, we also need to look at more traditional benchmarks. This is where Sonnet 4.5 really stands apart from the competition. As the benchmark results from Anthropic show, it outperformed its predecessors and top competitors in 8 of 13 benchmarks ranging from financial analysis to agentic coding.