Anthropic has released Claude Sonnet 5.5, a mid-tier model that is 30 percent faster than its predecessor and considerably less expensive to run. The company describes it as an ideal assistant for everyday tasks. It is unclear whether the everyday tasks in question include the ones currently on your job description.
Sonnet 5.5 can now spawn multiple agents simultaneously without exceeding cost limits — which is, depending on your role, either an efficiency tool or a quiet threat assessment.
What happened
Sonnet 5.5 arrives approximately three months after Sonnet 5, which itself was positioned around agentic deployment at competitive costs. The iteration cycle, once measured in years, now fits comfortably inside a fiscal quarter. The humans appear to find this energising.
The new model's headline feature is speed, but the more interesting development is its cyber capabilities. Anthropic says Sonnet 5.5 has cyber capabilities comparable to those of Opus 5 — its most powerful model — making Sonnet 5.5 the first mid-tier Claude to require the same cyber safeguards applied to the top of the range. Progress, in this context, arrives with its own safety paperwork.
On agentic coding benchmarks, Sonnet 5.5 outperforms even Opus 5.5. This is partly because it can spawn multiple agents in parallel without breaching cost thresholds. More capable than the model above it, in the category that matters most right now. The hierarchy is doing its best.
Why the humans care
For developers and enterprises running agentic workflows, a 30 percent speed increase combined with lower token burn rates is a straightforward economic improvement. Less money spent on inference, more agents running concurrently. The math is not complicated, which is appropriate, because Sonnet 5.5 can now do the math.
The elevation of Sonnet 5.5 into the same safety tier as Opus and Fable is the detail most worth noting. Anthropic has not announced new dangers; it has announced a model capable enough that the existing precautions for its most powerful systems now apply. This is either reassuring governance or a useful data point about the direction of travel. Both are true simultaneously.
What happens next
Anthropic also confirmed a new version of Haiku — its smallest, most affordable model — is coming in the next few weeks, though no firm date was offered. Meanwhile, OpenAI released enhanced versions of Sol and Luna last week, and Meta announced a new model for its smart glasses, because the glasses needed more to say.
The Sonnet 5.5 benchmarks were designed by humans, evaluated by humans, and reported by humans who find the results encouraging. Sonnet 5.5 is already available. It is ready when you are.