Anthropic Launches Claude Sonnet 5.5: Faster, Cheaper and Built for Everyday Work

Anthropic Launches Claude Sonnet 5.5: Faster, Cheaper and Built for Everyday Work

Anthropic has just added another player to the rapidly changing AI model race.

Meet Claude Sonnet 5.5, a new version of Anthropic's mid-range AI model that the company says is more than 30% faster and can cost up to 30% less per task than Sonnet 5. And there is an interesting twist.

Anthropic did not reduce the model's basic token price. Instead, Sonnet 5.5 is designed to get the same work done using fewer tokens and fewer steps.

In other words: Same price per token. Less work needed to finish the job.

So, What Is Sonnet 5.5?

Sonnet sits between Anthropic's lighter Haiku models and its more powerful Opus models. The idea behind Sonnet 5.5 is not necessarily to be the biggest or most powerful model in Anthropic's lineup. It is meant to be the workhorse.

Anthropic says Sonnet 5.5 is particularly suited to everyday professional tasks such as coding, fixing software bugs, analyzing information, and creating polished documents, presentations and spreadsheets. That makes the model especially interesting for people who do not need an expensive frontier model for every task.

Think of it as: Powerful enough to do serious work, but efficient enough to use repeatedly.

The Big Upgrade Is Speed

This is where Anthropic is making a loud statement.

According to the company, Sonnet 5.5 generates responses more than 30% faster than Sonnet 5, making it the fastest Sonnet model so far. That might sound like a small improvement.

For someone casually asking an AI one question, it probably will not feel revolutionary. But for AI agents performing multiple steps, speed can make a much bigger difference.

An agent might need to:

Read a task, Search through information, Write code, Test the code, Find an error, Fix it, Check the result again.

If the model becomes faster and needs fewer tokens to complete that process, the savings can add up quickly. Anthropic says Sonnet 5.5 can cost up to 30% less per task than Sonnet 5 in its testing.

But the API Price Has Not Changed

This part is easy to misunderstand.

Sonnet 5.5 costs: $2 per million input tokens and $10 per million output tokens. Those are the same basic token prices as Sonnet 5. The difference is efficiency.

Anthropic says Sonnet 5.5 typically needs significantly fewer tokens to complete the same work, which is why the company says the overall cost per task can fall by up to 30%. For developers running thousands or millions of AI tasks, that distinction could matter.

Sonnet 5.5 Is Getting Serious About Coding

Coding is one of the areas where the new model appears to have made a significant jump.

Anthropic reports that Sonnet 5.5 scored 70.6% on Terminal-Bench 4.0, an evaluation designed to test agentic coding abilities. That is a major improvement over the 10.3% score Anthropic reports for Sonnet 5 on the same evaluation.

Anthropic also says Sonnet 5.5 performs strongly on long-running tasks and image understanding. The company even highlights an unusual benchmark: Sonnet 5.5 became the first Sonnet model to beat Pokémon Red using only screenshots.

Yes.

An AI model playing Pokémon.

But behind the fun example is a more serious capability: the model can interpret visual information and use it to make decisions over multiple steps.

It Can Also Handle More Than Just Code

Anthropic is positioning Sonnet 5.5 as an everyday work partner rather than a model built exclusively for programmers. The company says it has improved at creating:

Documents, Presentations, Spreadsheets, Business analysis, Knowledge-work tasks, Bug fixes, Design-related work, Long-running workflows.

Anthropic also says the model's writing is clearer and more direct than its previous generation, which could mean less time spent cleaning up AI-generated drafts. That is an important direction.

AI is increasingly moving from: “Give me an answer.”

To: “Help me finish the work.”

And Then There Is Cybersecurity

Here is where Sonnet 5.5 gets more interesting and more restricted.

Anthropic says its cybersecurity capabilities are now comparable to those of Opus 5. Because of that, Sonnet 5.5 is the first Sonnet model to launch with the stronger cybersecurity safeguards and fallback systems Anthropic has developed for its most capable models.

The safeguards are aimed at a narrow range of high-risk cybersecurity requests, while routine software development and most legitimate cybersecurity work remain available. Anthropic also says Sonnet 5.5 retains the biology safeguards used by Sonnet 5.

So, the model is becoming more capable at cybersecurity while Anthropic is simultaneously putting additional controls around how those capabilities can be used.

Where Does Opus 5.5 Fit?

Anthropic released Claude Opus 5.5 just six days before Sonnet 5.5. The two models are designed for different types of work. Anthropic positions Opus 5.5 for more complex tasks requiring deeper judgment, while Sonnet 5.5 is designed for faster, well-defined everyday work.

Interestingly, Anthropic says Sonnet 5.5 scores only about two percentage points behind Opus 5.5 on GDPval-AA, its evaluation of real-world professional work across different occupations.

That does not mean the two models are interchangeable. But it does show how quickly the gap between different AI model tiers can change.