1300 166 138 Tell us what you need
News / Claude

Claude Sonnet 5.5 is here: what changed and who gets it

By Monika Tantau • Published 2 months ago
Claude Sonnet 5.5 is here: what changed and who gets it

Key takeaways

• Anthropic released Claude Sonnet 5.5 on 28 September, more than 30% faster than Sonnet 5
• Anyone can use it on Claude.ai, and developers get it on AWS, Google Cloud and Azure
• API price stays at $2.81 in and $14.04 out per million tokens, half of Opus 5.5
• At the same index score, Opus 5.5 costs $4.86 per task and Sonnet 5.5 costs $10.67

Most Read

Anthropic released Claude Sonnet 5.5 on 28 September, the second model in its Claude 5.5 family after Opus 5.5. According to Anthropic, it runs more than 30% faster than Sonnet 5 and costs up to 30% less per task, while the price per token stays exactly where it was.

The pitch is a division of labour. Opus 5.5 is for complex work that needs careful judgment over a long stretch; Sonnet 5.5 is for well scoped everyday tasks, fixing bugs and producing polished documents, slides and spreadsheets. For Australian teams paying for Claude by the token, the useful question is how much of their work the cheaper model can now carry.

What changed from Sonnet 5

The jump is largest in agentic coding. On Terminal Bench 4.0, which tests multi step work in a command line, Sonnet 5.5 scores 70.6% against 10.3% for Sonnet 5. On GDPval AA, a test of real work across 44 occupations, it lands two points below Opus 5.5 and about 400 above its predecessor. On OSWorld, a computer use test, it moves from 57.0% to 80.1%.

Most of the saving comes from using fewer tokens rather than a lower rate. Balyasny Asset Management told Anthropic the model used about 121,000 tokens per answer on its finance tasks, where Sonnet 5 used 497,000. Zendesk said support tickets were processed 20% faster. Anthropic still says Opus 5.5 is clearly stronger at complex, open ended work.

The price and the catch

API pricing is unchanged: $2.81 per million input tokens and $14.04 per million output tokens, converted at the Reserve Bank rate of 22 September. That is half of Opus 5.5, at $5.62 and $28.08, and the same per token price as OpenAI’s GPT 6 Sol. Anthropic bills in US dollars, so what a local team pays moves with the exchange rate.

On the Artificial Analysis Intelligence Index, Sonnet 5.5 at max effort scores 56. That matches Opus 5.5 at xhigh, trails only Opus 5.5 at max, and beats Claude Fable 5.1 on 53.4 and GPT 6 Sol on 47.5. The catch is the cost of getting there: at that score, Sonnet 5.5 runs about $10.67 per task and Opus 5.5 about $4.86, because Sonnet burns far more tokens at its top setting.

Lower down the scale the picture flips. At high effort, Sonnet 5.5 scores 46.7 for about $1.52 per task, while Opus 5.5 at medium scores 51.2 for about $1.88. For routine work Sonnet is the cheaper choice; for the hardest problems, Opus at xhigh gives the same quality for less than half the cost.

Who gets it and where

Who gets it and where

The short answer: everyone. Anthropic’s Sonnet page says anyone can chat with Claude using Sonnet 5.5 on Claude.ai, on the web and in the iOS and Android apps, which includes the free plan. Developers can call it today on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure. In the apps and in Claude Code the default effort is medium; on the Claude Platform it is high.

Anthropic has not said whether Sonnet 5.5 replaces Sonnet 5 as the default model for free and Pro users. It is also the first Sonnet to ship with the cyber safeguards used on Opus 5.5. Routine bug fixing is unaffected, but higher risk security requests visibly fall back to Sonnet 5, so security teams may notice the switch.

Why Australian businesses should care

Most business AI work sits in Sonnet territory: drafting replies, summarising documents, pulling figures out of reports and building first cuts of slides. Because the per token price has not moved, switching to Sonnet 5.5 should lower the bill for those jobs, but only if the prompts are tested first. Our guide to choosing a Claude model sets out where each tier fits.

Zero data retention is available on Sonnet 5.5, as it is on Opus 5.5. Under the Privacy Act 1988 the organisation stays accountable for personal information it sends to a provider, so that option matters for customer service and finance workflows. Teams running Sonnet with thinking switched off need to move to the new between_tools setting before migrating, which is worth checking in a chatbot build before it breaks in production.

Haiku 5.5 is next

Anthropic says Claude Haiku 5.5, aimed at high volume and cost sensitive work, will join the family in the coming weeks. Once it lands, the three tiers will be priced and tuned for different jobs, and the choice between them becomes a routing decision rather than a loyalty one. Our comparison of Opus 5.5 and Fable 5.1 shows how that routing works at the top end.

Scroll to Top