Anthropic Releases Claude Sonnet 5.5: 30% Faster, 30% Cheaper Than Sonnet 5

Anthropic has released Claude Sonnet 5.5, replacing Sonnet 5 as its medium-tier model. The company says the new model runs more than 30% faster and costs up to 30% less for most work compared to Sonnet 5, which launched in June.

What's new

Sonnet 5.5 is the second model in Anthropic's Claude 5.5 family, arriving one week after Opus 5.5. Anthropic had promised to upgrade both Sonnet and Haiku "in the coming weeks" when it released Opus 5.5.

According to Anthropic, Sonnet 5.5 performs best on everyday tasks with clear scope, including bug fixes and producing polished documents, slides, and spreadsheets. The company also highlights design capabilities, including refining user interfaces and generating slides from templates that require minimal editing. Anthropic says early testers found it a better collaborator than Sonnet 5, citing clearer writing and speed suited to quick iteration on less complex tasks.

The model is already live in Claude Code, according to posts from Anthropic staff on X.

Pricing and performance

Sonnet 5.5 keeps Sonnet 5's pricing structure:

• Input tokens: $2 per million
• Output tokens: $10 per million
• Cache reads: $0.20 per million

Anthropic claims the model typically uses fewer tokens to complete equivalent work, and generates output more than 30% faster than Sonnet 5 — making it, according to the company, the fastest Sonnet model released to date. No independent benchmark scores were disclosed in Anthropic's announcement.

Safety changes

Sonnet 5.5 is the first Sonnet-tier model to ship with cybersecurity safeguards and fallback protections comparable to those built into Opus 5, Anthropic's top-tier model. The company says these controls target a narrow set of high-risk requests and don't affect routine software development work.

Biology-related safeguards remain unchanged from Sonnet 5, and Anthropic says most life sciences use cases are unaffected.

What this means

Sonnet 5.5 is a speed-and-cost optimization rather than a capability leap — Anthropic isn't claiming new benchmark territory, just faster inference and lower effective token spend at unchanged list pricing. That's consistent with how the company has iterated its mid-tier model in the past: incremental efficiency gains between larger architectural updates.

The more notable shift is the safety posture. Extending Opus-grade cybersecurity safeguards down to the Sonnet tier suggests Anthropic sees meaningful misuse risk in a model priced for high-volume, everyday use — not just its flagship. That's a signal worth watching as competitors decide whether similar guardrails belong on mid-tier, cost-sensitive models or only on top-of-line releases.

For developers already on Sonnet 5, the upgrade path looks low-friction: same pricing, same API surface, faster responses. The real test is whether the claimed 30% speed and cost improvements hold up under independent benchmarking, since none of Anthropic's performance claims in this announcement come with published third-party evaluation numbers.