Anthropic has released Claude Sonnet 4.6, and the benchmark numbers are remarkable: 79.6% on SWE-bench Verified, placing it within just 2 percentage points of Opus. For a model in the Sonnet tier — faster and cheaper than Opus — that level of coding performance is a significant milestone.
In developer preference testing, Sonnet 4.6 was chosen over its predecessor Sonnet 4.5 roughly 70% of the time. The improvements are most noticeable in multi-file reasoning, instruction following, and the accuracy of code generation in unfamiliar frameworks. Developers report fewer hallucinated APIs, better understanding of project context, and more reliable refactoring suggestions.
Sonnet 4.6 is now the default model on Free and Pro plans, which means the majority of Claude users get an immediate upgrade without changing anything. For the COR community, this is particularly relevant because many members use Sonnet as their daily driver for faster tasks while reserving Opus for complex architectural work. The gap between the two has narrowed to the point where Sonnet can handle tasks that previously required Opus.
