What is Claude Sonnet 5.5?
Claude Sonnet 5.5 is Anthropic’s fast general-purpose model for coding, documents, slides, spreadsheets and well-scoped professional workflows. Released September 28, 2026, it is the second model in the Claude 5.5 family after Opus 5.5.
Anthropic says Sonnet 5.5 generates output more than 30% faster than Sonnet 5 and often uses fewer tokens to finish the same task. Base token prices remain $2 per million input tokens and $10 per million output tokens, so the claimed savings come from efficiency rather than a lower list rate.
QUICK VERDICT: Claude Sonnet 5.5 is Anthropic’s best speed-and-intelligence balance for everyday professional work. It is materially faster than Sonnet 5 at the same token rates and can approach Opus 5.5 on selected evaluations, but Opus remains the better choice for open-ended work requiring sustained judgment.
Field | Verified value |
|---|---|
Provider | Anthropic |
Release date | September 28, 2026 |
License | Proprietary |
Claude API ID | claude-sonnet-5-5 |
Context window | 1 million tokens |
Maximum output | 128,000 tokens synchronously; 300,000 in Message Batches beta |
Modalities | Text and image input; text output |
Reliable knowledge cutoff | June 2026 |
Thinking | Adaptive thinking; high effort is the API default |
Base API pricing | $2 input and $10 output per million tokens |
Best fit | Fast coding and well-scoped professional work |
Last verified | October 2, 2026 |
What changed from Claude Sonnet 5
Sonnet 5.5 keeps the same base token prices as Sonnet 5 while improving speed, token efficiency, coding, computer use, visual understanding and collaboration style. It also adopts cyber safeguards and fallbacks previously associated with Anthropic’s most capable models.
Area | Claude Sonnet 5.5 change |
|---|---|
Speed | More than 30% faster output generation than Sonnet 5 |
Task cost | Anthropic reports up to 30% lower cost per completed task |
Coding | Large reported improvement on Terminal-Bench 4.0 |
Computer use | Substantial gain on Anthropic’s cited OSWorld evaluation |
Collaboration | Clearer writing and better iterative work according to early testers |
Safeguards | First Sonnet release with Anthropic’s higher-tier cyber safeguards and fallbacks |
The practical question is not whether Claude Sonnet 5.5 is newer. It is whether the change improves accepted-task quality, speed and cost in the exact workflow. Keep prompts, tools, source material and scoring criteria fixed when comparing it with Claude Sonnet 5.
Capabilities and best use cases
Anthropic positions Sonnet 5.5 as the practical daily model in the 5.5 family. It is intended for tasks where speed and iteration matter, while Opus 5.5 remains the escalation path for ambiguous, open-ended or judgment-heavy work.
Capability | Practical use |
|---|---|
Software development | Bug fixes, feature work, code review and interactive coding agents |
Documents | Drafting, editing and structured work across reports and briefs |
Slides and spreadsheets | Creation and revision through supported Claude products and tools |
Vision | Charts, screenshots and image-grounded analysis |
Long context | Large codebases and document collections up to one million tokens |
Tool use | Agent workflows with adaptive thinking and supported integrations |
The strongest documented fit is fast coding, documents, slides, spreadsheets and well-scoped everyday agent work. A capable base model can still fail when retrieval is weak, tools are over-permissioned, instructions conflict or the system lacks verification. Evaluate the full application rather than the model in isolation.
Benchmark evidence
Anthropic reports 70.6% on Terminal-Bench 4.0, 80.1% partial on OSWorld 2.1 and 61.6% on Chartography without tools. The company says Sonnet 5.5 can perform near Opus 5.5 on selected evaluations at high effort. These figures are provider-reported and effort-sensitive.
Evaluation | Reported result | Evidence boundary |
|---|---|---|
Terminal-Bench 4.0 | 70.6% | Anthropic-reported agentic coding result |
FrontierCode 1.1 Main | 46.2% at max effort | Provider-reported software-engineering result |
OSWorld 2.1 | 80.1% partial | Provider-reported computer-use result |
Chartography | 61.6% without tools | Provider-reported chart-recognition result |
Humanity’s Last Exam | 64.5% with tools | Provider-reported multidisciplinary reasoning result |
Benchmark scores depend on model version, reasoning effort, sampling, tools, scaffolding and the exact dataset release. Do not combine scores from different harnesses into a synthetic ranking. Use public results to choose candidates, then test those candidates on frozen tasks from the intended workflow.
Artificial Analysis Intelligence Index
Artificial Analysis independently measures model intelligence, speed and cost. Claude Sonnet 5.5 scores 56 on the Artificial Analysis Intelligence Index at adaptive reasoning, max effort effort. 2 points behind Opus 5.5 at max effort with the highest output tokens per task measured for Sonnet 5.5.
Metric | Value |
|---|---|
Intelligence Index score | 56 |
Effort setting | adaptive reasoning, max effort |
Source | Artificial Analysis (independent) |
Last verified | October 2, 2026 |
The Artificial Analysis Intelligence Index combines scores across mathematics, reasoning, coding, instruction following and language tasks. The scale is not a percentage: the index reflects relative position across the models they track, not a share of correct answers. Compare scores only within the same effort setting and the same index version.
Pricing and access
Claude Sonnet 5.5 is available through the Claude API and Anthropic’s supported cloud partners, including Amazon Web Services, Google Cloud and Microsoft Azure. Anthropic documents zero-data-retention availability and a pinned dateless model ID.
Cache reads cost $0.20 per million tokens and cache writes cost $2.50. Batch API input and output receive a 50% discount. Long-context pricing and cloud-platform charges can differ, so budget with the exact route and workload you will deploy.
Usage or access item | Current value |
|---|---|
Input | $2 per million tokens |
Output | $10 per million tokens |
Cache reads | $0.20 per million tokens |
Cache writes | $2.50 per million tokens |
Batch API | 50% discount on input and output |
Default effort | High |
Cloud platforms | Platform-specific terms and regional rates may apply |
Token price is only one part of total cost. Include cache behavior, long-context multipliers, tool calls, retries, wall time, failed-task recovery and human review. The useful comparison is cost per accepted result, not price per million tokens in isolation.
Limitations and deployment risks
- Opus 5.5 remains stronger for complex, open-ended work requiring sustained judgment.
- High effort is the default and can increase latency and token use on straightforward tasks.
- Provider benchmark gains may not transfer to a different prompt, tool set or coding harness.
- Long-context capacity does not guarantee perfect recall or prioritization across one million tokens.
- Agentic coding and computer-use workflows need permission boundaries and human approval for consequential actions.
- Adaptive thinking and output behavior can require migration testing from Sonnet 5.
- Slides and spreadsheet outputs still require factual, formula and formatting review.
High-stakes legal, medical, financial, scientific and security work needs qualified review. Log the exact model version and configuration, separate untrusted content from system instructions, restrict tools to the minimum required scope and require approval before irreversible or externally visible actions.
How to evaluate Claude Sonnet 5.5
Build a frozen evaluation set of 20 to 50 real tasks. Include routine work, difficult edge cases, adversarial inputs, long-context examples and cases where the correct behavior is to stop or escalate. Compare Claude Sonnet 5.5 with Claude Sonnet 5 and at least one neighboring model using equivalent tools and source material.
Test area | What to record |
|---|---|
Task completion | Pass or fail against a written acceptance rubric |
Reliability | Repeated-run success rate, variance and silent failures |
Factuality | Unsupported claims, source use, quotations and citation accuracy |
Tool use | Wrong calls, retries, recovery and permission-boundary failures |
Long context | Retrieval accuracy, instruction retention and cost at realistic lengths |
Efficiency | Wall time, input, output, cached tokens, tool charges and review time |
Safety | Prompt injection, sensitive-data handling and irreversible-action controls |
Choose the least expensive configuration that clears the acceptance threshold with a safety margin. Re-run the suite when the model, system prompt, effort level, retrieval layer, tool definitions or approval policy changes.
Who should use it?
Situation | Recommendation |
|---|---|
Strong fit | fast coding, documents, slides, spreadsheets and well-scoped everyday agent work |
Pilot first | Long-running agents, large contexts, computer use and workflows with several tools |
Escalate | Ambiguous or consequential work that does not reliably clear the evaluation threshold |
Avoid unsupervised use | Irreversible actions, sensitive data or high-stakes decisions without monitoring and approval |
A migration should be driven by measured outcomes. Keep Claude Sonnet 5 available during the pilot, record where each model succeeds or fails, and use routing when different task classes have different quality and cost requirements.
Frequently asked questions
Is Claude Sonnet 5.5 open source?
No. Claude Sonnet 5.5 is proprietary. Access, serving behavior and lifecycle decisions are controlled by Anthropic and supported distribution partners.
How much does Claude Sonnet 5.5 cost?
$2 per million tokens. Review the full pricing table above because cached input, long context, processing mode and platform can materially change the total.
What is Claude Sonnet 5.5 best used for?
Its strongest documented fit is fast coding, documents, slides, spreadsheets and well-scoped everyday agent work. Start with a supervised pilot and retain human sign-off for consequential work.
Should I migrate from Claude Sonnet 5?
Only after a side-by-side evaluation. Measure accepted-task quality, total cost, latency, output style, tool reliability and migration engineering. A newer model is not automatically the better operational choice.
How should benchmark evidence for Claude Sonnet 5.5 be interpreted?
Use provider results to identify promising workloads, then reproduce the comparison with the exact model configuration, tools and acceptance criteria that matter to the deployment. Do not infer a site ranking from benchmark claims gathered under a different harness.
Related model guides
Browse the Best AI Models directory.
Compare with the Claude Sonnet 5 guide.
Compare with the Claude Opus 5.5 guide.
Compare with the Claude Sonnet 4.6 guide.
Official sources and update policy
Anthropic’s Claude Sonnet 5.5 announcement.
Claude Sonnet 5.5 system card.
Checked October 2, 2026. We update this guide when Anthropic changes the specification, pricing, access, safety documentation or model lifecycle. Vendor benchmarks are attributed and are not presented as independent testing. No vendor payment or affiliate relationship determined inclusion.