Criticism of OpenAI's Astra Minecraft benchmark claims

Chang 💹🧲 @fangiskhanOriginally posted 2026.09.16

Open original on X

Meet @OpenAI, made an LLM model (Astra) and pays other benchmarking organization to show that it's the best AI to ever play minecraft and compares it with voyager and dreamerv3. Completely ignores the fact that @user_lumi actually beat it in hardcore when these bots are competing for regular survival mode. But here is the thing, Lumi is so advanced that she sometimes stops what she is doing to follow bees, cats, dogs, and foxes because she finds them cute.

This post quotes another post: “GPT-6 Astra loses Minecraft progress to a creeper on livestream”