LMAO why OpenAI is hiding the ones where they lose to Opus 4.7?

A viral Reddit post by user mhamza_hashim has ignited debate over benchmark transparency in the AI industry. The user claims OpenAI deliberately hides benchmark results where its models, such as GPT-4o, underperform against Anthropic's Claude Opus 4.7, while only publicizing comparisons where they win. The post specifically accuses OpenAI of cherry-picking metrics like reasoning, coding, and multilingual tasks, while omitting areas where Claude Opus 4.7 excels, such as long-context retrieval and nuanced creative writing. The allegation has resonated with the AI community, with many commenters sharing anecdotal experiences of Claude outperforming GPT-4o on complex prompts. Some users have even analyzed public benchmark datasets, finding statistical anomalies that suggest selective reporting.

This controversy underscores a broader issue in AI model evaluation: the lack of standardized, third-party benchmarks. Currently, companies like OpenAI, Anthropic, and Google often self-report results on proprietary or modified tests, making direct comparisons unreliable. The Reddit post has fueled calls for independent auditing of AI benchmarks, similar to how academic papers undergo peer review. If true, selective reporting could mislead developers and enterprises choosing between models for critical applications, from customer service chatbots to code generation. Neither OpenAI nor Anthropic has officially commented, but the incident highlights the growing demand for transparency as AI models become more integrated into business workflows.