Anthropic launches Claude Sonnet 5.5 with near-Opus performance at half the price
The New Stack Frederic Lardinois ● Covered by 3 sources
Anthropic just shipped Claude Sonnet 5.5. It’s faster and cheaper, and keeps getting close to the pricier Opus model.
Based on reporting by The New Stack, Frederic Lardinois — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic on Monday released Claude Sonnet 5.5, the newest version of its workhorse model and the second release in the Claude 5.5 family. Opus 5.5 arrived last week. A new Haiku is also on the way, with Anthropic saying that smaller, cheaper model should land in the coming weeks.
Sonnet is meant to be the practical one, and Anthropic is selling this update as a big step forward on speed and cost. The company says it produces output more than 30% faster than Sonnet 5 while cutting cost per task by up to 30%. Anthropic also says it is the fastest Sonnet model yet, and positions it as a lower-cost companion to Opus 5.5.
The strongest gains are in coding. On Terminal-Bench 4.0, Sonnet 5.5 scores 70.6%, up sharply from 10.3% for Sonnet 5 and ahead of Opus 5.5’s 66.4%. On CursorBench, it lands within about two points of Opus 5.5. On Cognition’s FrontierCode, it scores 52.1% at its second-highest effort setting, versus 54.4% for Opus 5.5 and 49.3% for OpenAI’s GPT-6 Sol. Anthropic also says the new model writes more clearly, and that it is good at everyday tasks, bug fixing, polished documents, slides, and spreadsheets.
The gap to Opus is still there, though. On Artificial Analysis’ GDPval-AA, which measures real-world knowledge-work tasks across 44 occupations, Sonnet 5.5 scores 1,844. That is just two points behind Opus 5.5, but Anthropic still says Opus remains clearly stronger for complex work that needs sustained judgment. Sonnet 5.5 also comes close on computer use and Humanity’s Last Exam, and Anthropic says it is the first Sonnet model to beat Pokémon Red using only screenshots.
Pricing stays put at $2 per million input tokens and $10 per million output tokens, with cache reads at $0.20 per million tokens. Anthropic says the model should still run cheaper in practice because it uses far fewer tokens than Sonnet 5. Sonnet 5.5 is live on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Azure. Developers who had thinking turned off will need to switch to a new between_tools setting.
My take — AI-written commentary, not fact-checked reporting
Anthropic is doing the obvious sensible thing here: push the middle model hard until it eats into the flagship’s job. That’s a lot healthier than the usual AI circus where every release is sold as the moon, the stars, and a better tax return. The real story is not the benchmark gap; it’s how quickly “good enough” keeps getting cheaper, which is bad news for hype merchants and very good news for everyone else.
Read more about this at: The New Stack