Okay I tried @GeminiApp new Argon model (Gemini-4-Argon) in making minecraft.
The pros
I've used the exact same prompt for every model I've tested this on, and this was the fastest by and cheapest far. It took around 15 minutes and only $1.72.
The cons
The output is better than GPT 6.1 Sol, but nowhere close to opus 5.5 or even sonnet. The GUI is way off, the water physics are broken, and the tools are super lazily done, and when you kill mobs, they explode like a piñata.
Overall, Opus 5.5 is still king even if it took an hour and $15.
https://minecraft-model-bench.eclinick.chatgpt.site
+7
