Anthropic just dropped Claude Sonnet 5, and the pitch is essentially “near-Opus brains at Sonnet prices.” The model, which went live on June 30, targets developers and businesses who want high-end AI coding performance without the high-end invoice.
The introductory API pricing tells the story: $2 per million input tokens and $10 per million output tokens. Those rates hold until August 31, 2026, after which they bump to $3 and $15, respectively.
What Sonnet 5 actually does differently
The headline improvement is in coding and what Anthropic calls “agentic task performance.” In English: the model is better at handling multi-step software engineering processes, the kind where previous AI models would lose the thread halfway through and start hallucinating function names that don’t exist.
Sonnet 5 comes with a 1 million token context window. That’s enough room to ingest entire codebases, lengthy documentation sets, or complex project specifications without needing to chunk everything into bite-sized pieces first.
