The Sonnet 5.5 release date was 28 September 2026. Anthropic shipped it one week after Opus 5.5 as the new default mid-tier Claude, at the same price as Sonnet 5 but, by Anthropic's numbers, more than 30% faster and up to 30% cheaper per task.
📺 Watch: Claude Opus 5.5 Agent OS is SCARY GOOD!
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
If you run Claude every day, in Claude Code, through the API or inside an agent stack like Hermes, this is the model most of your work will land on. Below you get the date, where the model sits in the Claude lineup, what Anthropic says changed, the full pricing and specs, Anthropic's reported benchmarks, the breaking changes developers need to handle, and a plain call on whether you should switch today.
Sonnet 5.5 Release Date and Where It Lands in the Lineup
Anthropic released Claude Sonnet 5.5 on 28 September 2026. It is the second model in the Claude 5.5 family, arriving exactly one week after Opus 5.5. It replaces Sonnet 5, which launched in June 2026, as the mid-tier Claude model. On Anthropic's official model page, Sonnet 5.5 carries the status "Active (latest)", and Anthropic commits to not retiring it sooner than 28 September 2027. So if you build on Sonnet 5.5 today, you have at least a year of guaranteed runway before you need to plan a migration.
Here is how Sonnet 5.5 fits against the rest of the current Claude lineup, using Anthropic's official table. Prices are per million input and output tokens.
| Model | Input / Output price | Speed | Notes |
|---|---|---|---|
| Fable 5.1 | $10 / $50 | Slower | Top of the range |
| Opus 5.5 | $4 / $20 | Moderate | Hardest reasoning work |
| Sonnet 5.5 | $2 / $10 | Fast | New default mid-tier, 1M context |
| Haiku 4.5 | $1 / $5 | Fastest | 200K context, cheapest |
The takeaway is simple. Sonnet 5.5 sits in the sweet spot: half the price of Opus 5.5, much faster, and built for the everyday jobs that make up most of your Claude usage. If you want the full picture on Opus as the default inside Claude Code, read Claude Code with Opus 5.5 as the default model.
Want to see how people are actually putting new Claude models to work the week they drop? Inside the AI Profit Boardroom, 3,000+ members share their setups, prompts and workflows, and you lock in $69/mo instead of the normal $110.
What Anthropic Says Is New in Sonnet 5.5
Anthropic positions Sonnet 5.5 around two headline claims. First, Anthropic reports more than 30% faster output than Sonnet 5. Second, Anthropic says Sonnet 5.5 comes in at up to 30% lower cost for most work, which comes from finishing tasks more efficiently rather than from a price cut, since the per-token price is unchanged.
According to Anthropic, Sonnet 5.5 is strongest at everyday tasks with a clear scope. That covers bug fixing, document creation, UI refinement, and slide and spreadsheet design. Anthropic also highlights clearer writing and speed suited to quick iteration, which is exactly what you want when you are going back and forth with Claude on a draft, a landing page or a small code change.
There is a safety angle too. Anthropic says Sonnet 5.5 is the first Sonnet model with cyber safeguards comparable to Opus 5. For teams that were holding back on Sonnet for security-sensitive work, that is worth noting.
Julian has not benchmarked Sonnet 5.5 himself yet. His hands-on take, with real tasks and real scores, goes on Goldie Bench once he has run it. Until then, every performance figure on this page is Anthropic's.
📺 Watch: OpenClaw Mission Control + NotebookLM Update + Gemini 3.1 Pro vs Claude Sonnet 4.6!
Sonnet 5.5 Pricing and Specs
Anthropic kept Sonnet 5.5 pricing identical to Sonnet 5. Here is the full breakdown from the official Claude model page:
- Input: $2 per million tokens
- Output: $10 per million tokens
- Cache reads: $0.20 per million tokens
- 5-minute cache writes: $2.50 per million tokens
- 1-hour cache writes: $4 per million tokens
- Batch API: 50% off
On the spec side, Sonnet 5.5 gives you a 1M-token context window and 128K max output tokens. On the Batch API, you can push output up to 300K tokens with a beta header. Adaptive thinking is on by default, with the default effort set to "high". The model takes text and images in and returns text out, and its knowledge cutoff is June 2026.
The model ID is claude-sonnet-5-5 on the Claude API, Google Cloud, Microsoft Foundry and Claude Platform on AWS. On Amazon Bedrock it is anthropic.claude-sonnet-5-5. Sonnet 5.5 was also live in claude.ai and Claude Code at launch, so you did not have to wait for a staged rollout.
📺 Watch: NEW Grok 4.2 Update VS Claude Sonnet 4.6
The Sonnet 5.5 Benchmarks (Anthropic's Numbers)
These are the benchmark results Anthropic reports for Sonnet 5.5. They are Anthropic's figures, not independent tests, so treat them as the vendor's starting point rather than the final word.
- Terminal-Bench 4.0: 70.6%, per Anthropic
- CursorBench 4.0: 55.5%, per Anthropic
- OSWorld 2.1: 80.1%, per Anthropic
- FrontierCode 1.1: 46.2% at max effort, per Anthropic
- Humanity's Last Exam: 64.5% with tools, per Anthropic
- GDPval-AA v2.1: 1844 Elo, per Anthropic
The spread tells you where Anthropic wants Sonnet 5.5 to shine: terminal work, computer use and real-world knowledge tasks. The OSWorld score in particular points at agent-style work where Claude drives a desktop. If you are wiring Claude into your own agents, the Agent OS guide shows how to plug a model like this into a working system.
Breaking Changes If You Build on the Claude API
If you are moving an app from Sonnet 5 to Sonnet 5.5, Anthropic's migration notes list several breaking changes. In plain English:
- Forced tool use now returns an error. If your code forces Claude to call a specific tool, you need to change that approach.
- Thinking blocks are tied to the model and conversation. You cannot carry them across to a different model or reuse them in another conversation.
- The older computer-use tool version is not accepted on the Claude API and Google Cloud. Upgrade to the current version.
- Certain older models are rejected as advisor-tool pairings. Check which model you pair with Sonnet 5.5 if you use the advisor tool.
- Text between tool calls now arrives inside thinking blocks. Apps that stream that text to users need a display setting, or you can switch off up-front thinking with the between_tools setting.
- Non-default sampling values return a 400 error. If you set a custom temperature or other sampling values, remove them.
None of these changes are hard to fix, but they will break a production Claude integration if you just swap the model name and ship. Test in staging first. If you use Claude for search work, the Claude SEO skill guide is a good place to test the new model on real tasks.
Not sure whether your business should be building on Claude at all, or which model to put where? Book a free strategy session and get a plan for your own setup.
Who Should Switch to Sonnet 5.5
For most people, the switch to Sonnet 5.5 is an easy yes. Here is how to think about it:
- Claude Code and API users on Sonnet 5: this is a straight upgrade at the same price. You get Anthropic's claimed speed and cost gains without paying more per token. Just handle the breaking changes above if you build on the API.
- People doing the hardest reasoning work: stay on Opus 5.5. Anthropic still positions Opus as the model for the toughest problems.
- High-volume, cost-sensitive jobs: Haiku 4.5 remains the cheapest option at $1 input and $5 output per million tokens, with a 200K context window.
- Hermes users: you can pick up Sonnet 5.5 through the Claude plugin's live model picker. See the Hermes Claude plugin guide for setup, and compare it with running Hermes with Claude Opus 5.5.
The practical move is to make Sonnet 5.5 your default Claude model for everyday work, then escalate to Opus 5.5 only when a task genuinely needs deeper reasoning. That mix keeps your costs down while you keep the quality where it counts.
Sonnet 5.5 FAQ
When was Claude Sonnet 5.5 released?
Anthropic released Claude Sonnet 5.5 on 28 September 2026, one week after Opus 5.5. It replaced Sonnet 5 as the mid-tier Claude model and will not be retired sooner than 28 September 2027.
How much does Sonnet 5.5 cost?
Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, the same as Sonnet 5. Cache reads are $0.20 per million, 5-minute cache writes are $2.50, 1-hour cache writes are $4, and the Batch API is 50% off.
Is Sonnet 5.5 in Claude Code?
Yes. Anthropic made Sonnet 5.5 available in Claude Code and claude.ai at launch, alongside the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS.
Get More From Sonnet 5.5
A new Claude model is only as useful as the workflows you put it into. If you want ready-made prompts, agent setups and a community testing new models like Sonnet 5.5 the week they launch, join the AI Profit Boardroom. You get 3,000+ members and a locked $69/mo rate.
Want a direct plan for using Claude to grow your business? Book your free strategy session today. The Sonnet 5.5 release date was 28 September 2026, and the sooner you switch, the sooner you get the speed.
