Anthropic has launched Claude Sonnet 4.6, featuring a million token context window and enhanced capabilities in programming and reasoning.
Anthropic has introduced an advanced version of its Sonnet AI model, named Claude Sonnet 4.6, which emphasizes programming expertise and instruction adherence.
This is Claude Sonnet 4.6: our most capable Sonnet model yet. It represents a comprehensive upgrade across various tasks, including coding and long-context reasoning, along with a 1 million token context window in beta.
— Claude (@claudeai) February 17, 2026
Accessible to both Free and Pro plan users, the new model boasts a context window that allows it to handle extensive content, such as codebases, contracts, and academic papers, all in one prompt.
Benchmark results indicate that Sonnet 4.6 has achieved new highs in evaluations like OSWorld for computer use and SWE-Bench for programming tasks. Its score of 60.4% in the ARC-AGI-2 benchmark positions it ahead of many competitors, surpassed only by Opus 4.6 and a few other advanced models.
Earlier this month, the company also updated its premier model, Claude Opus, to version 4.6, enhancing its capabilities in action planning and efficiency with lengthy code tasks.
“Performance that previously required an Opus-level model is now available in Sonnet 4.6,” the company noted on its blog.
Additionally, Anthropic has reported securing $30 billion in funding, raising its valuation to $380 billion. This capital will support ongoing research and infrastructure development.
A safety evaluation of Sonnet 4.6 confirms significant improvements in its resistance to prompt injections compared to its predecessor.
In a related note, media outlets recently reported that the AI tool Claude was utilized in a military effort to apprehend Venezuelan President Nicolás Maduro.
