Claude Sonnet 5.5
API ID claude-sonnet-5-5
Launched September 28, 2026
Claude Sonnet 5.5 is an Anthropic model, released September 2026. It has a 1,000,000-token context window and up to 128,000 output tokens. It takes image input and supports reasoning. Standard price: $2.00 input and $10.00 output per 1M tokens.
- Status
- Active
- Released
- September 2026
- Shutdown
- not before Sep 28, 2027
- Replacement
- —
claude-sonnet-5-5 is active. Anthropic says it won't be retired before September 28, 2027.
Benchmarks
Scores as each source published themComparing with Claude Sonnet 5, the previous model in this line. A score is compared only with the same benchmark from the same source, at the same setting where both have it. Full comparison: prices, limits, features
Arena
| Arena | Claude Sonnet 5.5 | Claude Sonnet 5 | Change | Rank | Date | Source |
|---|---|---|---|---|---|---|
TextBy category
| 1475 ±8 | 1461 ±4 | +14 | 35 of 415 6,254 votes | Oct 13, 2026 | source |
| Text: Coding | 1527 ±15 | 1520 ±6 | +7 within range | 26 of 410 1,506 votes | Oct 13, 2026 | source |
WebDevOther settings
| 1774 ±14 | 1540 ±6 | +234 | 3 of 142 2,858 votes | Oct 13, 2026 | source |
| Vision | 1266 ±14 | 1263 ±7 | +3 within range | 39 of 165 2,068 votes | Oct 9, 2026 | source |
| Agent | 0.125 ±0.031 | 0.047 ±0.017 | +0.078 | 6 of 54 7,446 sessions | Oct 14, 2026 | source |
Rated as claude-sonnet-5.5-xhigh (xhigh effort); Claude Sonnet 5 as claude-sonnet-5-high (high effort). Some were rated at different settings; ratings on one arena share one scale, so the difference still reads directly. Text, WebDev and Vision ratings come from people voting between two anonymous answers; Text and Vision are style-controlled, as on arena.ai. Agent is Arena's separate score for real agent sessions, on its own scale. Not rated on Search and Document, where Claude Sonnet 5 is. Data: Arena (arena.ai), CC BY 4.0.
Epoch AI
| Benchmark | Claude Sonnet 5.5 | Claude Sonnet 5 | Change | Setting | Run | Source |
|---|---|---|---|---|---|---|
| FrontierMath Tier 4 | 80.5% | 29.3% | +51.2 pts | max effort | Sep 29, 2026 | source |
| FrontierMath Tiers 1-3 | 88.8% | 65.6% | +23.2 pts | max effort | Sep 29, 2026 | source |
| GPQA Diamond | 95.6% | 80.3% | +15.3 pts | max effort | Sep 29, 2026 | source |
| OTIS Mock AIME 2024-2025 | 100.0% | 80.0% | +20.0 pts | max effort | Sep 29, 2026 | source |
| SimpleQA Verified | 46.5% | 33.7% | +12.8 pts | max effort | Sep 29, 2026 | source |
Run independently by Epoch AI. Data: Epoch AI, Capabilities & benchmarking (epoch.ai/benchmarks), CC BY.
On cloud platforms
All platform dates| Platform | Platform ID | State | Shutdown | Time left | Replacement | Source |
|---|---|---|---|---|---|---|
| Amazon Bedrock | anthropic.claude-sonnet-5-5 | Active | not before Sep 28, 2027 | — | source |
Read daily from each platform's own pages. The shutdown date and countdown on this page use Anthropic's own date.
Features
Compare models| Feature | Value | Since | Source |
|---|---|---|---|
| Prompt cachingA prompt prefix you send again and again (system prompt, documents, tools) is cached, so later requests cost less and start faster. | Supported | in place when tracking began (October 6, 2026) | source |
| Batch APISend many requests as one job that runs in the background, usually finished within 24 hours, at a lower price than live requests. | Supported | in place when tracking began (October 6, 2026) | source |
| Image inputImages such as screenshots, photos and charts can go in the prompt for the model to read. | Supported | in place when tracking began (October 6, 2026) | source |
| Reasoning / thinkingThe model can work through a problem step by step before it answers: usually better answers to hard questions, at the cost of more tokens and time. | Supported (thinking: adaptive; default effort: high) | in place when tracking began (October 6, 2026) | source |
| temperatureSets how random replies are: lower is more predictable, higher is more varied. Accepted: it works. Ignored: the request succeeds but the setting has no effect. Rejected: the request fails with an error. | Rejected | in place when tracking began (October 6, 2026) | source |
| top_pLimits each next word to the most likely choices, another way to make replies more or less varied. Accepted: it works. Ignored: the request succeeds but the setting has no effect. Rejected: the request fails with an error. | Rejected | in place when tracking began (October 6, 2026) | source |
| Context windowThe most the model can work with in one request, in tokens: your prompt, conversation history and files. A token is about three-quarters of an English word. | 1,000,000 tokens | in place when tracking began (October 6, 2026) | source |
| Max output tokensThe longest reply the model can write in one response, in tokens. A token is about three-quarters of an English word. | 128,000 tokens | in place when tracking began (October 6, 2026) | source |
Not published by Anthropic: Structured outputs, Function calling, PDF input, Fine-tuning, Built-in web search, logprobs.
Price history
USD per 1M tokens, standard tierWe started recording this model's prices on Oct 1, 2026. The history grows with each daily check.
Show as a table
| From | Type | Price |
|---|---|---|
| Oct 1, 2026 | Input | $2.00 |
| Oct 1, 2026 | Output | $10.00 |
Prices
USD per 1M tokens, every type and band, by tier| Type | Price $/1M | In effect |
|---|---|---|
| Standard | ||
| input | $2.00 | |
| cached input | $0.10 | |
| cache write (5 min) | $2.50 | |
| cache write (1 hour) | $4.00 | |
| output | $10.00 | |
| Batch | ||
| input | $1.00 | |
| output | $5.00 | |
Changelog
Changes we detected on Anthropic's pages- Price changed
Anthropic cut claude-sonnet-5-5 price: cached input $0.20 → $0.10 per 1M tokens
Breaking
Related news
- PricingBreakingAnthropic changed claude-sonnet-5-5 standard cached input price: $0.2 -> $0.1 per 1M tokens
- SDKsInfov1.9.0
- ReleasesBreakingWe've launched Claude Sonnet 5.5 (claude-sonnet-5-5).
Email alerts
Get an email if claude-sonnet-5-5 changes
We'll email you the same morning we spot a deprecation notice, a new or moved shutdown date, a price change, or the model leaving Anthropic's model list.
No password: one email to confirm, and you can unsubscribe any time. Privacy
Last checked · Sources: Anthropic models, Anthropic deprecations, Anthropic pricing, Anthropic model pages