Google Gemini 4 Argon agents defeat Ender Dragon in Minecraft

A team of four agents completed the task in 27 hours and 14 minutes after 605 deaths.

Google Gemini 4 Argon agents defeat Ender Dragon in Minecraft
Photo: Mashable

A team of four Gemini 4 Argon agents defeated the Ender Dragon, the final boss of Minecraft, without human help. The agents started in a freshly generated survival world on Easy difficulty with the goal to kill the dragon. The run took 27 hours and 14 minutes.

The AI players died 605 times during the run, mostly to monsters. Google DeepMind researchers said this is the first time AI agents have beaten Minecraft from start to finish using only standard tools and no game-specific help built by humans.

Researchers watched the run on a live dashboard and could have stepped in at any point, but they say they never needed to. However, the game ran on Easy, which meant monster attacks dealt less damage to the AI players.

The agents knew what blocks were hidden behind other blocks, making the mining part of the game rather insignificant. Because the agents play through code rather than on a screen, they have access to information that a human player would not. Additionally, the agents knew the optimal strategies for playing the game given the Gemini 4 Argon training base.

The run cost about $520 in model usage at the introductory API price for Argon. Researchers argue the skills involved, such as working as a team, bouncing back from mistakes, and sticking with a goal for more than a day, are the same skills AI agents need for real-world work.

Making a frontier AI model play a video game is a common benchmark used by many companies when testing.

Key facts

  • Four Gemini 4 Argon agents defeated the Ender Dragon in 27 hours and 14 minutes.
  • The AI players died 605 times, mostly to monsters.
  • The run cost approximately $520 in model usage.

Three perspectives

Neutral

Google DeepMind researchers reported that Gemini 4 Argon agents defeated the Ender Dragon using standard tools on Easy difficulty.

Positive

Researchers say the skills used, such as teamwork and persistence, are the same skills AI agents need for real-world work.

Negative

The AI agents had access to information a human player would not, such as knowing what blocks were hidden behind other blocks.

Your week in news, every Wednesday

The biggest stories of the week for your edition, in your inbox every Wednesday morning, your time. Free, and you can unsubscribe at any time.

Choose your news coverage

Please choose and confirm your coverage.

All sections

Edition