Brain the Size of a Planet: Are LLMs Thonking too Hard? (30 minute read) (opens in new tab)

Earlier studies showed that Opus 4.6, GPT 5.4, Gemini 3.1-pro-preview, Deepseek R1-0528, and Qwen 3.6-plus were unable to find two of the vulnerabilities discussed in the Mythos blog post without extremely revealing hints. This study continues the previous experiments with 26 distinct Claude-4.6/4.7 and GPT-5.4/5.5 combinations and different context window sizes and reasoning efforts. It found that higher reasoning effort, and even later models, are not always better for triaging security res...