Released: 2026-07-03
google/gemini-3-flash-preview
4
7
8
2
9
1
5
6
3
5
2
9
7
3
6
4
1
8
1
3
6
4
5
8
7
9
2
7
4
1
5
6
3
8
2
9
2
9
4
8
1
5
6
3
7
3
8
5
9
7
2
1
5
4
6
5
3
1
4
7
9
8
2
9
1
7
3
2
4
8
5
6
8
2
1
6
8
9
3
4
7
openai/gpt-4o
openai
7
6
9
8
2
1
5
3
4
5
4
2
9
3
7
6
1
8
1
3
8
4
5
6
7
9
2
3
7
1
5
6
4
8
2
9
2
9
4
8
1
3
7
5
6
6
8
5
2
7
9
4
1
3
4
1
3
7
9
8
2
6
5
9
7
6
3
4
5
1
8
7
8
2
7
6
8
2
3
4
7
meta-llama/llama-3.3-70b-instruct
meta
5
8
9
4
2
1
6
3
7
2
4
6
8
3
5
7
9
1
1
3
7
9
6
4
2
5
8
8
5
2
1
9
7
4
6
3
4
1
3
6
5
8
9
7
2
9
7
5
3
1
2
8
4
6
6
2
4
7
8
3
1
5
9
3
9
1
5
4
6
7
2
8
7
6
8
2
9
5
3
1
4
mistralai/mistral-medium-3
mistral
4
7
6
9
2
1
5
3
8
5
9
2
7
3
6
4
1
8
8
3
1
4
5
9
6
2
7
7
4
3
5
6
2
8
9
1
2
9
4
8
1
3
7
6
5
6
8
5
1
7
4
9
3
2
1
5
3
2
9
8
6
7
4
9
2
7
3
4
5
1
8
6
3
6
8
6
8
7
2
4
9
deepseek/deepseek-chat-v3-0324
deepseek
4
8
6
2
9
1
5
7
3
5
7
2
6
3
9
4
1
8
1
3
9
4
5
7
2
6
0
7
1
3
5
6
4
8
9
2
2
9
4
8
1
3
6
5
7
6
8
5
9
7
2
1
3
4
8
6
3
7
4
5
9
2
1
9
4
1
3
2
8
7
0
6
0
2
0
6
8
0
3
4
7
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.