Claude Sonnet 5.5: what's new, what it costs and how it compares with Opus 5.5

Claude Sonnet 5.5 is Anthropic’s newest Sonnet model, released on September 28, 2026, six days after Claude Opus 5.5. Anthropic calls it “a clear upgrade over Claude Sonnet 5” that runs more than 30% faster and costs up to 30% less for most work. This report uses Anthropic’s announcement, its model documentation and its release notes, and the benchmark figures are Anthropic’s own.

Claude Sonnet 5.5 at a glance

What it is for

Anthropic positions Sonnet 5.5 as a faster, lower-cost complement to Opus 5.5. Opus 5.5 is for complex work that needs careful judgment, while Sonnet 5.5 is strongest at well-scoped everyday tasks, fixing bugs and creating polished documents, slides and spreadsheets. Testers highlighted its design sense and its writing, which Anthropic says is clearer than the previous generation’s, and in one internal test two experts judged its first draft of a 10-slide operating review, built from earnings materials and a slide template, ready to send as is.

It is also the first Sonnet to beat Pokémon Red working only from screenshots, which Anthropic uses to illustrate its long-horizon and image-understanding gains.

Benchmarks

Anthropic’s table compares Sonnet 5.5 with its predecessor, Opus 5.5 and OpenAI’s GPT-6 Sol. These are vendor-reported results.

The pattern is that Sonnet 5.5 jumps well past Sonnet 5 and lands within a few points of Opus 5.5 on most tests. On Terminal-Bench 4.0 it scores higher than Opus 5.5’s best result, though Anthropic’s chart for that test uses GPT-5.6 Sol as the OpenAI comparison because OpenAI didn’t report GPT-6 Sol on it. Anthropic also notes that on FrontierCode Sonnet 5.5 scored lower at Max effort (46.2%) than at Xhigh, because it more often ran a code-review step that split work across many subagents and, in two cases examined, timed out or made edits beyond the task’s scope.

Anthropic adds a caution: benchmark scores capture only one facet, and in its own testing and external testers’ Opus 5.5 “remains clearly stronger at complex, open-ended work requiring sustained judgment”. It says Sonnet 5.5 complements Opus best at lower effort settings, where on several tests Low or Medium effort beats Sonnet 5’s best score for about a tenth of the cost per task.

What it costs

The price is unchanged from Sonnet 5, at $2 per million input tokens and $10 per million output tokens, with cache reads at $0.20 and 50% off in batch. The saving comes from tokens: Anthropic says Sonnet 5.5 typically needs far fewer tokens for the same work, so it costs up to 30% less per task, and it generates output more than 30% faster, making it the fastest Sonnet so far. Early-tester quotes in the announcement back this up, such as Slack reporting about 14% fewer output tokens on its Slackbot evaluations. Here is how the current lineup compares on price:

Our guide to Claude models explained covers the whole lineup and when to use each one.

Safety

Because Sonnet 5.5’s cybersecurity capabilities are comparable to Opus 5’s, it is the first Sonnet model to launch with cyber safeguards and fallbacks like those on Anthropic’s most capable models. Its biology safeguards are the same as Sonnet 5’s. Anthropic says both target a narrow set of high-risk requests and that routine software development and most life sciences work are unaffected. On its automated behavioural audit it improves on or matches Sonnet 5 on most measures of alignment. Sonnet 5.5, like Opus 5.5 and Sonnet 5, is available with zero data retention.

What developers need to change

Moving from Sonnet 5 isn’t a one-line change. Anthropic’s documentation lists five breaking changes:

1. Turning off up-front thinking now uses a new between_tools setting. If you run Sonnet with thinking off, you need to switch to it.
2. Forced tool use returns an error.
3. Thinking blocks are tied to the model and the conversation that produced them.
4. On the Claude API and Google Cloud, the earlier computer_20251124 computer-use tool isn’t accepted.
5. The advisor tool rejects Claude Opus 4.8, Opus 4.7 and Sonnet 5 as advisors.

One more change alters the response shape without failing a request: text between tool calls comes back in thinking blocks, so an app that streams that text to users goes quiet between tool calls until it sets a display value or turns off up-front thinking. Setting temperature, top_p or top_k to a non-default value returns a 400 error. Read the migration guide before switching production traffic. If you code with Claude, see how to use Claude Code and the best AI coding assistants.

Using it in the Claude apps

Anthropic’s announcement says Medium effort is the default in the Claude apps. The pages we read don’t list Sonnet 5.5’s availability plan by plan in the apps, so check the model picker in your account. Anthropic’s Claude Code documentation lists claude-sonnet-5-5 as a supported model. For how the plans differ, see Claude Pro vs Max, and for the assistant overall, the Claude profile.

How it compares with OpenAI’s latest

OpenAI’s GPT-6.1 Sol, announced the day after Sonnet 5.5, is priced at the same $2 and $10 per million tokens. The vendors publish different benchmark sets, so neither launch post settles which is better for your work, and it is worth testing both on your own tasks.