New Delhi, Feb. 28 -- It's evident how Sonnet can be used on Claude 3.7. This supports extended reasoning, and a user can control how much time the model spends with a question in terms of tokens. Here are the following benchmark results:

* SWE-Bench Score (GitHub issue solving): 62.3% (70.3% with extended thinking)

* Multi-Lingual Understanding Benchmark: 83.2% (86% with extended thinking)

* Outperforms GPT -4 in agentic and coding tasks.

Sonnet will also be available on the Claude app starting Monday as well as Anthropic API, Amazon Bedrock and Google Cloud's Vertex AI for only $3 per million input tokens and $15 per million output tokens, which is a higher rate than GPT-40 Mini.

In other news, Anthropic AI has hired Niki Parmar, a...