# Claude Sonnet 5.5: What It Does, Pricing and How to Use It

By Schnalz.net · 29 September 2026

https://schnalz.net/ai-guides/claude-sonnet-5-5-what-it-does-pricing-and-how-to-use-it

> Claude Sonnet 5.5 is Anthropic's new mid-tier model, launched on 28 September 2026. It runs 30%+ faster than Sonnet 5, costs up to 30% less per task at the same list price, and comes close to Opus 5.5 on most benchmarks.

## In short

- Sonnet 5.5 is the second model in Anthropic's Claude 5.5 family, after Opus 5.5.
- It runs 30%+ faster than Sonnet 5 and costs up to 30% less per task.
- List prices stay at $2 per million input tokens and $10 per million output tokens.
- It is available on the Claude Platform, AWS, Google Cloud, Microsoft Azure and in Claude Code.

## What Claude Sonnet 5.5 Is

Claude Sonnet 5.5 is the second model in Anthropic's Claude 5.5 family, following Opus 5.5, which launched the week before. Anthropic pitches it as the quicker, cheaper partner to Opus 5.5. Opus takes on the hardest, most open-ended problems; Sonnet 5.5 is aimed at clearly defined day-to-day jobs such as bug fixes, reports, presentations and spreadsheets, and Anthropic says it has "a sharp eye for design."

**Claude Sonnet 5.5 at a glance**

- **Developer**: Anthropic
- **Released**: 28 September 2026
- **Model tier**: Mid-tier (between Haiku and Opus)
- **Input price**: $2 per million tokens
- **Output price**: $10 per million tokens
- **Cache reads**: $0.20 per million tokens
- **Speed vs Sonnet 5**: 30%+ faster output generation
- **Cost vs Sonnet 5**: Up to 30% less per task (fewer tokens needed)

Sonnet is the mid-tier model in Anthropic's Claude family, between the smaller Haiku and the larger Opus, and Anthropic also has more capable models such as Fable and its Mythos-class models. Haiku 5.5, the smallest and fastest model in the family for high-volume, cost-sensitive work, is due in the coming weeks.

Like Opus 5.5, Sonnet 5.5 writes more clearly and in a more natural-sounding voice than previous-generation Claude models. SiliconANGLE reports that it also carries invisible text watermarking. When Anthropic introduced the technology, SiliconANGLE writes, it said the watermark would not hurt text quality or readability but would make AI-generated text easier to detect, and the feature was added to comply with global rules such as the EU's AI Act.

## Benchmark Performance

Sonnet 5.5 shows large gains over Sonnet 5 and comes close to Opus 5.5 on several evaluations. Its biggest improvements are in coding tasks.

**Benchmark scores (– means the sources give no clear figure; Sonnet 5.5's FrontierCode score is at its second-highest effort setting)**

|  | Sonnet 5.5 | Sonnet 5 | Opus 5.5 | GPT-6 Sol |
| --- | --- | --- | --- | --- |
| Terminal-Bench 4.0 (agentic coding) | 70.6% | 10.3% | 66.4% | – |
| FrontierCode (code mergeability) | 52.1% | – | 54.4% | 49.3% |
| CursorBench (real coding sessions) | 55.5% | – | 57.8% | – |
| GDPval-AA (real-world work) | 1,844 Elo | 1,449 Elo | 1,846 Elo | 1,487 Elo |

Anthropic's charts also show the savings: on several benchmarks, Sonnet 5.5 set to low or medium effort scores higher than Sonnet 5 at its best, while each task costs around a tenth as much. In a less formal test of long tasks and image understanding, it also became the first Sonnet model to finish the game Pokémon Red with only screenshots to go on.

> **Note**: Despite these scores, Anthropic says Opus 5.5 is still the better model when a problem is complex, loosely defined and needs careful judgment over a long stretch.

## What Early Testers Report

Several companies shared results from early testing. These give a practical sense of what the model can handle.

- Zendesk reported that Sonnet 5.5 processed support tickets 20% faster than the Claude models it uses in production, with fewer wrong decisions.
- Epic said it met the quality bar of a higher-tier model in its tests, worked across tens of thousands of lines of gameplay architecture code and kept going on tasks that lasted hours, with less detailed prompting.
- Unity reopens each project and tests the result at runtime before a task counts as done, and most of Sonnet 5.5's work got through. The model also completed 90% of the tasks in Unity's multi-step Editor and coding benchmark.
- In one tester's 118 real app builds, Sonnet 5.5's apps scored as well as Opus 5's. It needed 3.6 rounds per build on average against Opus 5's 7.7, and no other model in the comparison had fewer failed tool calls.
- In side-by-side runs, testers saw it group more tool calls into each step than Sonnet 5 did, which cut both the number of steps and the cost.

## Where to Access Claude Sonnet 5.5

Sonnet 5.5 is available now through the following channels:

- Claude Platform
- Amazon Web Services
- Google Cloud
- Microsoft Azure
- Claude Code (confirmed live at launch)

> **Watch out**: If your app calls Sonnet with thinking switched off, change it to the new "between_tools" setting before you upgrade to Sonnet 5.5, The New Stack reports. Opus 5.5 does not accept requests with thinking fully off.

## Pricing Comparison

Sonnet 5.5 keeps Sonnet 5's list price, which The New Stack says is also the list price of OpenAI's GPT-6 Sol. Anthropic says the savings come from the model needing far fewer tokens for the same work, which in its testing cut the cost per task by up to 30%.

**Claude Model Pricing (per million tokens)**

| Model | Input | Output | Cache Reads |
| --- | --- | --- | --- |
| Claude Sonnet 5.5 | $2 | $10 | $0.20 |
| Claude Opus 5.5 | $4 | $20 | Not stated |

The New Stack notes that Anthropic recently cut Opus 5.5 to $4 per million input tokens and $20 per million output tokens, and that OpenAI halved its prices for GPT-6 Sol and Luna on the same day. Per token, Sonnet 5.5 still costs half as much as Opus 5.5.

## Safeguards and Limits

Because Anthropic considers Sonnet 5.5's cybersecurity capabilities comparable to those of Opus 5, it is the first Sonnet model to ship with the same class of cyber safeguards and fallbacks used for Anthropic's most capable models. Higher-risk cybersecurity requests fall back to Sonnet 5. Biology safeguards remain unchanged from Sonnet 5.

Anthropic says these safeguards only apply to a small group of high-risk requests, so everyday software work and most life sciences research are not affected.

According to The New Stack, it is also the first Sonnet model with classifiers meant to block attackers who try to harvest its reasoning as training data for models of their own.

> **Tip**: For complex, open-ended tasks that need deep reasoning over long stretches, Opus 5.5 is still the stronger choice. Sonnet 5.5 is best suited to well-defined tasks where speed and cost matter most.

## Frequently asked questions

### How much does Claude Sonnet 5.5 cost?

Sonnet 5.5 is priced at $2 per million input tokens, $10 per million output tokens and $0.20 per million cache reads. These rates are the same as Sonnet 5, but Anthropic says the model typically uses far fewer tokens per task, making it up to 30% cheaper per task in its testing.

### What is the difference between Claude Sonnet 5.5 and Opus 5.5?

Sonnet 5.5 is faster, costs half as much per token and comes close to Opus 5.5 on most benchmarks. Anthropic says Opus 5.5 remains "clearly stronger at complex, open-ended work requiring sustained judgment", while Sonnet 5.5 is aimed at clearly defined everyday work such as bug fixes and documents.

### Where can I use Claude Sonnet 5.5?

Sonnet 5.5 is available on the Claude Platform, Amazon Web Services, Google Cloud, Microsoft Azure and Claude Code. If you use Sonnet with thinking switched off, move to the new "between_tools" setting before you upgrade.

### Is Claude Sonnet 5.5 good for coding?

Coding is where Sonnet 5.5 shows its largest improvements. It scored 70.6% on Terminal-Bench 4.0 (up from Sonnet 5's 10.3%) and came within about two points of Opus 5.5 on CursorBench. Early testers said it gets to grips with a codebase quickly and needs fewer steps than Sonnet 5 because it groups tool calls together.

### When will Claude Haiku 5.5 be released?

Anthropic says Haiku 5.5, the smallest and most cost-efficient model in the Claude 5.5 family, will launch in the coming weeks after Sonnet 5.5. No specific date has been announced.

## Sources

- [Anthropic: Introducing Claude Sonnet 5.5](https://www.anthropic.com/claude-sonnet-5-5)
- [SiliconANGLE: Anthropic debuts Claude Sonnet 5.5 running 30% faster than the previous-generation AI model](https://siliconangle.com/2026/09/28/anthropic-debuts-claude-sonnet-5-5-running-30-faster-than-the-previous-generation-ai-model/)
- [The New Stack: Anthropic launches Claude Sonnet 5.5 with near-Opus performance at half the price](https://thenewstack.io/claude-sonnet-55-launch/)
- [9to5Mac: Anthropic upgrades Claude with new Sonnet 5.5 model, details here](https://9to5mac.com/2026/09/28/anthropic-upgrades-claude-with-new-sonnet-5-5-model-details-here/)