Anthropic released Claude Sonnet 5.5 on Monday, six days after launching Opus 5.5, and for most people the smaller model may turn out to be the more important release of the two.
Sonnet is the model built for the work people bring to Claude every day, like rewriting an email, summarizing a long document, turning messy notes into something usable or building a slide deck.
Anthropic says Sonnet 5.5 generates responses more than 30% faster than Sonnet 5, writes more clearly and handles well-defined everyday tasks better than its predecessor. On one of the benchmarks that best reflects that kind of work, it scores almost exactly the same as Opus 5.5.
That combination has me rethinking which Claude model I start with.
Why speed matters for everyday Claude use
Most of my time with Claude isn't spent on hard problems. I might ask it to summarize something, explain a complicated topic, help plan a project or work through a problem, and then follow up as the answer takes shape.
That back-and-forth is where a faster model earns its keep, because a few extra seconds of waiting on every response can quickly add up when you're having a longer conversation.
The speed boost comes with an efficiency gain as well. Anthropic says Sonnet 5.5 needs far fewer tokens to finish the same job, so while developers pay the same $2 per million input tokens and $10 per million output tokens as they did for Sonnet 5, the company says the cost of a typical task drops by up to 30%. Opus 5.5 costs twice as much per token.
The gap with Opus has nearly closed on paper
The number that really caught my attention comes from a test designed to see how well AI models handle the kinds of work people do in the real world. It covers tasks from 44 different occupations, so rather than testing obscure trivia or math problems, it’s meant to give us a better idea of how useful these models might actually be at work.
Sonnet 5.5 scored 1,844. Opus 5.5, Anthropic’s more powerful model, scored 1,846. That’s just two points separating them. The previous version of Sonnet scored 1,449, so this is a pretty big jump. On this particular test, at least, the cheaper Sonnet has gone from trailing Opus by a wide margin to essentially tying it.
Sonnet 5.5 even came out ahead on one coding test, although Opus still won on others. That doesn’t mean Sonnet is suddenly the better model across the board, and Anthropic isn’t claiming that it is.
Opus 5.5 is still designed for the really difficult stuff: complicated projects, ambiguous problems and work that requires the model to keep track of a lot of moving pieces over a long period of time.
But most people aren’t asking Claude to solve those kinds of problems every day. And these results suggest that for the more ordinary things we use AI for, choosing Sonnet may no longer feel like much of a compromise.
When I'd still reach for Opus 5.5
Opus 5.5 is still the top model in the Claude 5.5 family, and I'd switch to it for a sprawling coding project, a tricky analysis or anything where Claude needs to keep track of a lot of interconnected information over a long session. For drafting an email, analyzing a document, brainstorming, working through research or building a presentation, I'd start with Sonnet 5.5.
That flips the way I've usually approached Claude. I used to wonder whether Sonnet was capable enough for a given task, and now I'd assume it is and only move up to Opus when a prompt proves otherwise.
How to get Claude Sonnet 5.5
Sonnet 5.5 is available now across Anthropic’s platforms, including the Claude API under the model name claude-sonnet-5-5, as well as through Amazon Web Services, Google Cloud and Microsoft Azure.
It’s available across Claude’s plans and replaces Sonnet 5 as the default model for Free and Pro users. In the Claude apps and Claude Code, it runs at Medium effort by default, which favors faster responses and lower token use, while developers using the Claude Platform get High effort by default.
Anthropic says a smaller Haiku 5.5 model, designed for high-volume and cost-sensitive tasks, will follow in the coming weeks.
5 prompts to try now
Here are five ways to get started with Claude Sonnet 5.5 now. You can tweak these prompts to fit your needs, just adjust what I've highlighted in bold:
Messy email: Rewrite this email so it sounds friendly but professional. Keep it under 120 words and don't make it sound overly formal: "Hi, just checking in because I sent this over last week and haven't heard anything. I know you're probably busy but I need to know where things stand because I have to make a decision by Friday. Can you let me know either way? Thanks."
Planning with real constraints: Plan a Saturday in New York City for a family of five with kids ages 5, 9 and 12. We have $200 to spend from 10 a.m. to 7 p.m., not including transportation. Include at least one indoor activity, somewhere casual for lunch and enough downtime that the day doesn't feel exhausting.
Explain something complicated simply: Explain why mortgage rates can stay high even after the Federal Reserve cuts interest rates. Assume I know nothing about economics. Keep it conversational, avoid jargon and give me one simple analogy that makes it easier to understand.
Make a decision rather than just list options: I'm buying a laptop mostly for web browsing, email, streaming, video calls and occasional photo editing. I tend to keep my computers for at least five years. Should I spend $700 on a basic model or $1,200 on a more powerful one? Walk me through the tradeoffs, then make a recommendation instead of telling me "it depends."
Handle an everyday problem with judgment: I have 45 minutes before guests arrive. My kitchen is messy, there are toys all over the living room, the bathroom needs cleaning, laundry is sitting on the couch and I haven't taken out the trash. Give me a realistic 45-minute plan. Prioritize what guests will actually notice and tell me what I should ignore until later.
Bottom line
The most powerful AI model isn't always the one I want to use. Most of my conversations with Claude aren't especially complicated. I want a good answer, I want it quickly and I want the model to understand what I'm asking without making me work for it. Sonnet 5.5 seems built for that kind of everyday use.
I still want to spend more time testing it side by side with Opus 5.5 before deciding just how close the two really are. But based on what I've seen so far, Sonnet 5.5 is probably the Claude model I'd reach for first.
Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds. Subscribe to Tom's Guide on YouTube and follow us on TikTok.
More from Tom's Guide
