Kimi K3.1 Passes the Rubik's Cube Test
Read the Portuguese translation →
Author: Matthew Berman · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Kimi K3.1 Passes the Rubik’s Cube Test · Channel: Matthew Berman · 00:31 · Watch on YouTube
Topics: AI evaluation and benchmarks, AI models
Summary
Section titled “Summary”Matthew Berman says he expected models to pass the Rubik’s Cube test easily, but notes that open-weight models have struggled with it. The cube is fully scrambled, and the model uses an algorithm to solve it rather than simply reversing the original moves. It calculates the required turns in advance and then plays them out in 3D. The successful solve earns an enthusiastic verdict: an absolute pass and one of the best.
Key points
Section titled “Key points”- The test checks whether a model can solve a fully scrambled Rubik’s Cube.
- The model solves the cube with an algorithm rather than reversing the initial moves.
- It pre-calculates the sequence of turns.
- The turns are played out in 3D space.
- The cube is solved successfully, followed by a confetti animation.
- Berman calls the result an absolute pass and one of the best.
- The video notes that open-weight models have not generally done well on this test.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | The Rubik’s Cube test | Berman introduces the test and says open-weight models have struggled, then explains that the cube is fully scrambled and will be solved algorithmically. |
| 00:10 | The model solves the cube | The model pre-calculates its turns and plays them out in 3D, successfully completing the solve. |
| 00:31 | Verdict | Berman calls the result phenomenal, an absolute pass, and one of the best. |
Quotes
Section titled “Quotes”I thought this test was dead. — 00:00
open weights models do not do well with the Rubik’s Cube test. — 00:00
This will actually use an algorithm to solve it. — 00:00
This is an absolute pass. In fact, one of the best. — 00:31
Concepts and entities
Section titled “Concepts and entities”Kimi K3.1, Rubik’s Cube test, Matthew Berman, open-weight models
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 I thought this test was dead. I thought all the models could pass it easily at this point, but it turns out open weights models do not do well with the Rubik’s Cube test. All right, so it is fully scrambled and it will solve it and not just reverse the set of moves. This will actually use an algorithm to solve it. So, let’s solve it. Here it is turn by turn. So, it actually pre-calculated all the turns it’s going to need and now it’s just playing it out in 3D space. But, there it is. We got a little confetti animation and this is
00:31 phenomenal. This is an absolute pass. In fact, one of the best.