Skip to content

Blog

Insights into my everyday programming life and other interesting technical tidbits

Posts about: ki, meinung Clear filter

April 2026

Published on Apr 17, 2026 · approx. 3 min read

Claude Opus 4.7: Benchmark Reality vs. Marketing Promises

Anthropic released Claude Opus 4.7 celebrating record benchmarks. A look at the independent numbers reveals: GPT-5.4 beats it in several categories, there are regressions compared to its predecessor, and the new tokenizer makes it effectively more expensive. A sober assessment.

Read More →