TechCrunch reported on Feb. 17 that Anthropic has released a new AI model, Sonnet 4.6.
The company said Sonnet 4.6 has improved coding and context understanding capabilities. It said the update was carried out in line with Anthropic’s strategy of improving models on a four-month cycle. Sonnet 4.6 will be offered as the default model to free and pro users.
Sonnet 4.6 expands the context window to 1 million tokens. The company explained that this is double the previous model and can handle large codebases, long contracts and multiple research papers in a single request.
TechCrunch said Sonnet 4.6 set new records at launch in OS World (computer use) and SWE-Bench (software engineering).
It scored 60.4 percent on the ARC-AGI-2 test and also stood out on indicators evaluating human intelligence-specialised capabilities. But it still falls short of Opus 4.6, Google Gemini 3 DeepSync and OpenAI GPT 5.2.
Anthropic recently unveiled Opus 4.6 and plans to update the Haiku model in the coming weeks.
Chi-gyu Hwang
delight@d-today.co.kr