Anthropic cuts Sonnet 5.5 cache-read price by half
Anthropic says reads of stored material cost $0.10 per million tokens, down from $0.20, while input and output rates stay at $2 and $10 per million tokens.
Anthropic says reads of stored material cost $0.10 per million tokens, down from $0.20, while input and output rates stay at $2 and $10 per million tokens.

Sole source: Anthropic. The company says its Claude Sonnet 5.5 model halves the price of reading cached material compared with Sonnet 5. Cache reads cost $0.10 per million tokens, down from $0.20; ordinary input and output prices remain unchanged. The company says the model also typically uses fewer tokens to do the same work.
Tokens are units used to measure the text a model processes or generates. A cache stores material for reuse, so the lower rate applies when that stored material is read again. Anthropic says its own tests found Sonnet 5.5 cost up to 30% less per task than its predecessor; that is separate from the fixed reduction in the cache-read rate.
The company positions the model for defined everyday tasks such as fixing bugs and making documents, slides and spreadsheets. It presents Sonnet 5.5 as a faster, lower-cost complement to Opus 5.5, which it says remains stronger at complex, open-ended work requiring sustained judgment.
In Anthropic’s evaluations, Terminal-Bench tests coding tasks and GDPval-AA tests work across 44 occupations and nine major industries. The company cautions that benchmark scores capture only one aspect of a model’s capabilities.
| Evaluation | Sonnet 5.5 | Sonnet 5 | Opus 5.5 |
|---|---|---|---|
| Terminal-Bench 4.0 | 70.6% | 10.3% | 66.4% |
| GDPval-AA v2.1 | 1,844 | 1,449 | 1,846 |

Effect Builder lets US residents over 18 create and preview custom Instagram Stories filters. Users can submit effects through the website, but a safety review must clear them before others can use them.
Anthropic says users who run Sonnet with thinking turned off must switch to the between_tools setting before moving to the new model; that setting keeps up-front thinking off. The company says higher-risk cybersecurity tasks will visibly fall back to Sonnet 5, while users can still find and fix bugs in routine software development.
Anthropic says cyberdefenders will soon be able to apply to its expanded Cyber Verification Program for tiered access to more advanced capabilities on Sonnet 5.5, Opus 5.5 and Claude Mythos models. It also plans to add Haiku 5.5, intended for high-volume and cost-sensitive applications, in the coming weeks.
1 min

Developers can use the new model in Muse Code and through the Meta Model API, with Meta positioning it for repeated tasks alongside the larger Muse Wonder.
3 min

The company also announced Claude Mythos 5.2 on 9 November, with access initially limited to vetted organizations.
1 min