Claude comes in three sizes. Opus is the heavy thinker, Haiku is the small fast one, and Sonnet sits in the middle as the everyday workhorse. It is the model most people actually use, whether they know it or not. Sonnet 5.5 arrived one week after Opus 5.5, and Haiku is the one still to come.
Anthropic's pitch is short: more than 30 percent faster than Sonnet 5, clearer writing, and exactly the same price. No new plan, no new bill. If you already use Sonnet, you get the upgrade for free.
A launch claim is a claim, so we ran our own test through the Anthropic API on launch day. Same prompt, same email, sent to both models at the same moment. Sonnet 5.5 finished about 35 percent faster.
Then something harder: build a complete website for a small bakery. Sonnet 5.5 finished in under 18 seconds, about four and a half seconds ahead of Sonnet 5, and the design it produced was the cleaner of the two. One prompt per race is not a lab study, so treat these as a real-world taste, not a verdict. Your results may vary.
Closer than a middle model usually gets. On real coding tasks Anthropic puts Sonnet 5.5 within about two points of Opus. It is also the first Sonnet to beat Pokémon Red while playing only from screenshots, which sounds like a party trick but is a decent test of an AI staying on task for a long time.
One more change worth knowing: Sonnet 5.5 is the first Sonnet with the same cyber safeguards as Anthropic's top models. In practice, a few security or biology questions may get a more careful answer than you are used to. That is a design choice, not a malfunction.
Here is where our test surprised us. On default settings, for that quick email, Sonnet 5.5 stopped to think for about seven seconds before it wrote a single word. On that run it finished slower than the old model.
That is not a bug. Thinking before answering is exactly what makes a model better at hard problems. It is just wasted time when all you want is a two line reply.
The fix is one setting. For quick jobs, set the effort to low, in the API or in Claude Code. For hard problems, leave it alone and let it think. That is where the extra time pays off.
Most small businesses do not need the biggest model. They need the one that is fast enough, cheap enough, and good enough to handle the same jobs every week: replies, summaries, first drafts, small bits of code. Sonnet 5.5 makes that everyday tier noticeably better, as long as you match the effort to the job. If you are still working out which jobs to hand to AI in the first place, start with why AI needs a second brain to be useful.
The whole update fits in one line: faster, clearer, same price, and one setting to know.
The free quiz takes 2 minutes and gives you a starting list for your business. No call, no card.
Take the free 2-minute quiz