Anthropic just released Claude Opus 4.5 yesterday. From what I'm seeing it's crushing the SWE-bench benchmarks – better than GPT-5 for software engineering tasks apparently.
Major highlights:
- Top tier at coding and computer use
- Better safety against prompt injection
- Supposedly outperforms humans on some tech interview problems
Anyone tried it yet? Wondering if it's worth switching from GPT for dev work or if the benchmarks are just marketing.
