Jiekou.AI
A third-party model guide for Claude Sonnet 4 (Anthropic's mid-tier model in the Claude 4 family, released May 22, 2025) reports it scored 72.7% on SWE-bench Verified at $3 per million input tokens, while Opus 4 cost five times more at $15 per million input for less than two additional percentage points of coding accur The guide describes Sonnet 4 as accepting text, images, and PDFs as input, supporting tool calling and multi-step task execution, with a 200K token context window and up to 64,000 tokens of extended internal reasoning during which it can call external tools and adjust its approach. Anthropic measured a 65% reduction in