Is this the first time OpenAI compared their new release to Anthropic models? Previously they were comparing only to GPT's own previous versions.