Back to today’s briefing

Story perspectives

Anthropic Launches Claude Sonnet 4.5: Top Coding Model

9/29/2025

50 8 Full Breakdown

Summary

Story summary

Anthropic unveiled Claude Sonnet 4.5, claiming it is the best coding model worldwide.
It can autonomously code for over 30 hours, up from seven hours.
It earned top scores on SWE-Bench Verified and OSWorld, surpassing GPT-5.
New tools include a VS Code extension and Claude Agent SDK.
Pricing remains $3 per million input tokens and $15 per million output tokens.

History

Historical Context

OpenAI's release of GPT-3 in 2021 marked a significant advancement in AI coding, utilizing a 175 billion parameter architecture. Following this, Anthropic launched its Claude models, starting with Claude in March 2023, which emphasized safety and efficiency in coding. These developments reflect a broader trend in AI technology reshaping software development practices and industry standards.

1 of 2

Story summary

Anthropic unveiled Claude Sonnet 4.5, claiming it is the best coding model worldwide.
It can autonomously code for over 30 hours, up from seven hours.
It earned top scores on SWE-Bench Verified and OSWorld, surpassing GPT-5.
New tools include a VS Code extension and Claude Agent SDK.
Pricing remains $3 per million input tokens and $15 per million output tokens.

1 / 2