Gemini 4 Argon 表现波动被吐槽像真人

Max @maxt3chno原推发布于 2026.10.17

在 X 上打开原推

Guys, Gemini 4 Argon feels more human than any other model WHY? because how tf else do you explain it dropping such different quality from task to task - one test it’s barely competing with Luna, Haiku - next test it’s beating Grok, Sol and Sonnet - third one it’s lowkey close to Opus 5.5 literally acts like a real person: - got drunk yesterday and slept 2h = output is trash - 5h of sleep = it’s mid - ate good and slept well = it cooks like crazy on this AWP Dragon Lore vibe check it's def closer to Opus 5.5 while being cheaper and faster