What Sonnet 5 is, what it costs, how it stacks up, what changed with Sonnet 5.5, and the one thing that matters more than the model you pick.
Anthropic released Claude Sonnet 5 on June 30, 2026. Then on September 28 it shipped Sonnet 5.5, which is now the current Sonnet. Sonnet 5 is still around as a legacy model, so here’s both.
But I want to say the quiet part out loud first. The model isn’t the thing that makes your AI good. More on that at the end, because it’s the part most people miss.
Claude comes in tiers. Haiku is the fast, cheap one. Sonnet is the everyday workhorse. Opus is the heavy hitter for long, complex work, and Fable now sits above it for the hardest reasoning jobs. (Anthropic’s model lineup)
When Sonnet 5 launched, Anthropic said it performs close to Opus 4.8, its top Opus at the time, at a much lower price, and called it "the most agentic Sonnet model yet," meaning it can plan, use tools and run multi-step tasks on its own. Anthropic was also clear that Opus 4.8 stayed the more generally capable model.
Both Sonnet 5 and 5.5 hold 1 million tokens of context, about 555,000 words, and think adaptively, spending more effort on hard questions and less on easy ones.
If you only use Claude in the app, none of this changes your bill. Your Free, Pro or Max plan stays the same. Pricing matters when you or your team build on the API, the pay-as-you-go version developers use.
01
Sonnet 5
$2 per million tokens in, $10 out. It launched with that as an intro price, and in August Anthropic made it permanent, so the planned jump to $3 / $15 never happened.
02
Sonnet 5.5
Same $2 / $10. Anthropic says it often needs fewer tokens, so a task can cost up to 30% less than on Sonnet 5. For comparison, Opus 5.5 is $4 / $20 and Fable 5.1 is $10 / $50.
One honest catch
Sonnet 5 counts tokens differently from Sonnet 4.6. The same text can come out to up to about 35% more tokens. The lower per-token price mostly cancels that out, but don’t assume every job got cheaper.
Anthropic calls Sonnet 5.5 "a clear upgrade" over Sonnet 5. It runs 30%+ faster and is much stronger at agentic coding, long multi-step work and understanding images. Early testers said it writes more clearly too.
Anthropic is also honest that Opus 5.5 is still clearly stronger at complex, open-ended work that needs sustained judgment. My rule: Sonnet 5.5 for everyday work and speed, Opus 5.5 when the task is big and complicated. If you’re still on Sonnet 5 in the API, switch the model ID to claude-sonnet-5-5.
Here’s what I actually care about. A new model drops every few weeks now, and Sonnet 5 itself got replaced in three months. Chasing each one is exhausting, and it’s not where the real advantage lives.
What makes your AI great is everything it knows about you, from your work and goals to the way you talk. And so many people keep all of that trapped inside one app. It’s like training the perfect assistant for a year, then watching everything they learned walk out the door when they quit.
Do the opposite. Keep it in plain files on your own device. That’s portable context. Your notes live in files any AI tool can read, whatever model or app you use.
01
Start your second brain in 10 minutes
Make a folder with three plain text files. about-me covers who you are and your goals. how-i-write holds your tone and a few real samples. projects lists what you’re working on now. Paste this prompt into any AI and it’ll draft all three with you.
Then paste or attach those files before you ask for anything, or drop them into a Claude Project’s files so every chat starts with them.
Once your context lives in files, you’re free. If Claude goes down, you switch in seconds. If another tool gets better at something, you bring your files and go. You never re-explain yourself, and you’re ready for the next launch the day it drops.