Sonnet 4.6 Max - unable to follow instructions to a T
Users are reporting that Anthropic's Claude Sonnet 4.6 Max model is failing to adhere to complex, multi-step instructions. Despite explicit prompts to process large text files section by section and confirm completion, the model appears to be sampling or truncating content, only reading a fraction of the provided text. This behavior is hindering users' ability to perform detailed analysis and summarization tasks, leading them to seek alternative methods like scripting or one-shot prompting to overcome the model's limitations. AI
IMPACT Highlights potential limitations in current LLM instruction following for complex tasks.