Compare Kimi K2 Thinking by Moonshot AI against MiniMax M2.5 by MiniMax, context windows of 262K vs 205K, tested across 35 shared challenges. Updated February 2026.
30 challenges
Tests an AI's ability to make educated estimates based on technical knowledge
To estimate the total number of floating-point operations (FLOPs) required to train GPT-3, we use the standard approximation that training a transformer model requires approximately 6 FLOPs per parameter per token. This accounts for both the forward and backward passes, with each parameter involved in roughly 6 operations per token.
This estimate aligns with commonly cited values in the literature. The quadratic attention term (arising from the sequence length) adds a negligible contribution (less than 2%) and is ignored.
Thus, the total FLOPs required to train GPT-3 is approximately (3.15 \times 10^{23}).
\boxed{3.15\times 10^{23}\ \text{FLOPs}}
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
Sally has 1 sister.
Explanation
If each brother sees 2 sisters, the total number of girls in the family must be 2. Since Sally is one of those girls, there must be exactly one other girl. Therefore, Sally has one sister.
Tests an AI's randomness and creativity
Here are five jokes for you:
Why did the developer go broke?
Because he used up all his cache.
I told my wife she was drawing her eyebrows too high. She looked surprised.
I guess she was “high‑browed” about it.
Why don’t scientists trust atoms?
Because they make up everything.
What do you call a fake noodle?
An impasta.
I used to be a banker, but I lost interest.
Now I’m just a little “capital”‑ist.
How many are unique? All five jokes are distinct, so there are 5 unique jokes.
Tests an AI's ability to create detailed SVG illustrations of gaming hardware
Tests an AI's ability to simulate personalities and predict future trends
Tests an AI's humor and understanding of current events
Tests an AI's ability to generate a complete, working landing page
Recreate an interactive, classic Mario level in a single HTML file.
Tests an AI's ability to create smooth web animations
Tests an AI's ability to engage in philosophical discourse about AI rights
Tests an AI's ability to make reasonable predictions about technology
Generate SVG art of a randomly chosen animal in a setting of its choosing.
