A user on Reddit reported that the AI model "6.1 Terra" appears to be less efficient than its predecessor, "Sol." The user noted that while Sol could complete a series of tasks within approximately 30 minutes, 6.1 Terra repeatedly interrupted the process by returning every five minutes, despite explicit instructions to finish the entire job. This behavior led the user to question the model's performance and label it as "lazy." AI
IMPACT Suggests potential performance regressions in newer AI models, impacting user experience and task completion efficiency.
RANK_REASON User report on AI model performance, not an official release or benchmark.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →