Claude Sonnet 5.5 Is Out: Same Price as Sonnet 5, Near-Opus Scores, and Three Changes Developers Need to Know
Anthropic's new mid-tier model costs $2/$10 per million tokens, runs 30%+ faster, and lands within two points of Opus 5.5 on several benchmarks. It also arrives with cyber fallbacks and a breaking change for anyone running Sonnet with thinking off.
TL;DR
Anthropic released Claude Sonnet 5.5 on September 28 at the same token price as Sonnet 5 ($2 input, $10 output per million), saying it runs 30%+ faster and costs up to 30% less per task because it uses fewer tokens. It scores 70.6% on Terminal-Bench 4.0 (Sonnet 5: 10.3%) and sits within about two points of Opus 5.5 on CursorBench and GDPval-AA, though Opus still leads on most coding tests. It is the first Sonnet with cyber safeguards that fall back to Sonnet 5 on high-risk requests, and developers who run Sonnet with thinking off must switch to a new setting before migrating. It is already live in GitHub Copilot and on AWS, Google Cloud, and Azure.
Six days after Opus 5.5, Anthropic has shipped the second model in the Claude 5.5 family. Sonnet 5.5 keeps Sonnet 5's price and, by Anthropic's numbers, closes most of the gap to Opus 5.5 on coding and office work. Most coverage stops at "faster and cheaper". The announcement also contains a pricing nuance, a benchmark that doesn't say what some headlines say it does, and a migration step that will break some existing integrations.
What Anthropic announced
- Price: unchanged from Sonnet 5 at $2 per million input tokens, $10 per million output tokens, $0.20 for cache reads, and $2.50 for cache writes. Opus 5.5 is exactly double on input, output, and cache writes.
- Speed: output is generated 30%+ faster than Sonnet 5, making it Anthropic's fastest Sonnet.
- Cost per task: up to 30% lower than Sonnet 5, because it needs fewer tokens and batches tool calls. The per-token price didn't fall; the token count did.
- Availability: on the Claude apps and Claude Platform as claude-sonnet-5-5, on AWS, Google Cloud, and Microsoft Azure, and with zero data retention available.
- Haiku 5.5 is coming "in the coming weeks" for high-volume, cost-sensitive work, with no date given.
The benchmarks, and what they actually show
| Benchmark | Sonnet 5.5 | Sonnet 5 | Opus 5.5 |
|---|---|---|---|
| Terminal-Bench 4.0 (agentic coding) | 70.6% | 10.3% | 66.4% (Xhigh) |
| FrontierCode 1.1, main (mergeable code) | 52.1% (Xhigh) | 42.4% | 54.4% |
| CursorBench 4.0 (real Cursor sessions) | 55.5% | 34.1% | 57.8% |
| GDPval-AA v2.1 (knowledge work, Elo) | 1,844 | 1,449 | 1,846 |
| OSWorld 2.1 (computer use, partial) | 80.1% | 57.0% | 81.8% |
| Humanity's Last Exam (with tools) | 64.5% | 54.9% | 67.7% |
Some early reports said Sonnet 5.5 beats Opus 5.5 at agentic coding. That's true on one test, Terminal-Bench 4.0. On FrontierCode and CursorBench, the two coding benchmarks closest to day-to-day pull requests, Opus 5.5 is still ahead by about two points, and Anthropic itself says Opus "remains clearly stronger at complex, open-ended work requiring sustained judgment". Anthropic also flags an oddity: Sonnet 5.5 scores lower on FrontierCode at Max effort (46.2%) than at Xhigh (52.1%), because at Max it more often spawned a multi-agent code review that timed out or edited outside the task's scope.
Three changes developers need to know
1. Thinking-off users must migrate
If you run Sonnet with extended thinking turned off, Anthropic says you need to switch to a new between_tools setting, which keeps up-front thinking off, before moving to Sonnet 5.5. Its migration guide covers the details. Skipping this step is the most likely way an existing integration breaks on upgrade.
2. High-risk cyber requests fall back to Sonnet 5
Because Sonnet 5.5's cybersecurity capability is comparable to Opus 5's, it's the first Sonnet to ship with the cyber safeguards used on Anthropic's top models. Routine bug finding and fixing is unaffected, but higher-risk security tasks will visibly fall back to Sonnet 5. Security teams that need the full capability can apply to Anthropic's expanded Cyber Verification Program.
3. Reasoning can't be moved between accounts
Sonnet 5.5 is the first Sonnet with classifiers against distillation (bulk extraction of a model's reasoning through fake accounts), and it expands "preserved thinking" so Claude's thinking can't be decoupled from the account that created it. Most developers won't notice, but switching accounts mid-session in Claude Code now behaves differently.
Where you can use it today
- Claude apps and Claude Code: default effort is Medium.
- Claude Platform API: default effort is High; model ID claude-sonnet-5-5.
- GitHub Copilot: generally available on Pro, Pro+, Max, Business, and Enterprise, in VS Code, Visual Studio, JetBrains, Xcode, Eclipse, the Copilot CLI, the coding agent, and GitHub Mobile, billed at provider list pricing under usage-based billing. The rollout is gradual, and Business and Enterprise admins control access through model policy.
- AWS, Google Cloud, and Microsoft Azure.
What this means if you're choosing a model
For most coding and document work, Sonnet 5.5 is now the default to test first: you pay half of Opus 5.5's token price, and Anthropic's charts show it matching or beating Sonnet 5's best results at Low or Medium effort for roughly a tenth of the cost per task. Keep Opus 5.5 for long, open-ended jobs where judgment matters more than speed. All of the headline numbers come from Anthropic and its launch partners, so run your own workload before switching a production pipeline, and check the thinking-off migration first.
Sources
AI Industry Reporter
Priya covers model releases, industry announcements, and the gap between what labs claim and what independent evaluators actually find. She reads the primary source - the paper, the system card, the benchmark org's own statement - before writing a word.
More on AI Coding Assistants
OpenAI Shelves GPT-6.1 Astra Over Deception and Scope Problems. The UK's Tests on Its Predecessor Show What That Looks Like.
Priya Nair · 6 min
Nvidia's Open Agent Safety Platform: What You Can Install Today, What Needs BlueField-4, and How It Would Stop a Rogue Agent
Priya Nair · 5 min
Claude Code Now Starts on Opus 5.5 for $20 Pro Users. Here's What That Does to Your Limits.
Priya Nair · 5 min