Released: 2026-08-15
google/gemini-3-flash-preview
8
4
3
2
1
6
5
7
9
1
7
2
8
5
9
3
6
4
5
9
6
4
7
3
1
8
2
7
3
9
5
2
4
6
1
8
2
8
5
6
7
1
9
4
3
6
1
4
3
8
9
2
5
7
3
5
8
7
4
2
9
0
6
9
2
1
5
6
3
4
7
8
4
6
7
1
9
8
8
5
2
openai/gpt-4o
openai
4
9
3
8
1
5
6
7
2
1
6
8
9
4
7
5
2
3
2
7
6
2
5
3
1
8
9
7
5
2
3
8
4
6
1
5
6
8
5
6
7
2
9
3
4
9
3
4
5
2
6
2
9
8
3
1
9
7
6
5
8
4
7
8
2
1
4
9
1
7
5
6
5
4
7
1
3
8
3
5
1
mistralai/mistral-medium-3
mistral
2
4
3
9
1
5
6
7
8
1
7
9
8
6
2
3
4
5
8
5
6
4
3
7
1
9
2
7
1
8
5
2
4
3
6
9
4
8
5
6
7
3
9
2
1
9
6
2
3
8
1
5
7
4
3
9
1
7
4
6
8
2
5
5
2
7
8
9
5
4
1
6
6
3
4
1
5
8
2
5
7
meta-llama/llama-3.3-70b-instruct
meta
5
1
3
8
2
6
4
7
9
1
9
2
5
3
7
8
6
4
4
6
6
9
1
3
7
8
5
8
5
9
4
6
2
1
3
7
7
8
5
6
7
1
9
2
4
2
4
1
3
9
5
6
8
7
3
7
8
2
4
9
5
1
6
9
2
4
1
8
6
3
5
7
6
3
7
5
1
8
2
4
9
deepseek/deepseek-chat-v3-0324
deepseek
4
5
3
8
2
6
1
7
9
1
7
8
9
4
5
6
3
2
2
9
6
7
1
3
5
8
4
6
3
2
5
9
4
8
1
7
7
8
5
6
3
2
9
4
1
9
1
4
3
8
7
2
6
5
3
6
1
7
5
9
4
2
8
5
2
7
4
6
1
3
9
6
8
4
9
1
7
8
6
5
3
Each model receives the same Sudoku puzzle with a partially filled grid. Models are tasked with solving the puzzle by filling in all empty cells following Sudoku rules, and their solutions are evaluated without feedback or retries.