The new benchmark king
Anthropic just shipped Claude 4.5 Sonnet with a 1 million token context window, native tool use, and what they're calling "extended thinking." Early benchmarks are wild — it's outperforming GPT-4o on every major coding eval.
The real story isn't the benchmarks though. It's the tool use. Claude can now chain together file reads, web searches, and code execution in a single turn. That changes the game for agentic workflows.
What this means for builders
If you're building AI-powered apps, the 1M context window means you can feed entire codebases into a single prompt. No more chunking, no more RAG gymnastics for small-to-medium projects.
The pricing is competitive too — roughly on par with GPT-4o for the output quality you get.