New model "beats GPT"? On a benchmark the vendor ran on itself. Nobody checked. Agentic Weekly labels every AI release Reported (sourced, not run) or Tested (we ran it, logs public). MCP server + open

原文：https://www.a8q9.cn/en/article/new-model-beats-gpt-on-a-benchmark-the-vendor-ran-on-itself-nobody-checked-agent-597a85.html