Teortaxes, identified in the post as a vocal supporter of DeepSeek, made the observation that recent advances in model post-training have matched the spread of actual applications. As a result, the latest benchmarks no longer indicate that models from labs other than Anthropic or OpenAI are underperforming significantly. The statement was shared publicly along with a reference to an external status update. This reflects the creator's view on how evaluation methods have evolved with industry developments. No independent verification of the benchmark results is provided in the available evidence.