Released: 2026-07-08
google/gemini-3-flash-preview
1
6
7
5
8
4
9
3
2
8
2
9
1
3
2
6
7
5
3
5
4
6
7
9
8
2
1
9
3
8
2
6
1
5
4
7
4
7
6
9
1
5
2
8
3
5
1
2
4
3
7
6
9
0
2
9
3
1
7
6
4
5
8
6
8
1
3
9
5
7
2
4
7
4
5
8
2
1
3
9
6
openai/gpt-4o
openai
1
6
5
7
8
4
9
3
2
8
4
9
1
3
2
6
7
5
3
7
2
6
5
9
8
1
4
5
3
7
2
9
8
1
4
6
4
9
6
5
1
3
7
8
2
6
1
8
4
2
7
5
2
3
2
8
3
9
7
1
4
5
8
9
2
1
3
6
5
2
6
7
7
5
4
8
3
6
9
9
1
meta-llama/llama-3.3-70b-instruct
meta
5
6
1
4
8
3
2
3
2
8
5
9
1
6
2
4
7
5
3
4
2
6
5
9
8
1
6
1
3
5
2
4
7
9
6
8
4
8
6
5
1
3
7
2
9
9
1
7
8
2
6
5
4
3
2
9
3
7
4
1
6
5
8
6
7
4
3
9
5
1
8
2
7
2
8
9
3
4
6
1
5
mistralai/mistral-medium-3
mistral
5
6
1
7
4
8
9
3
2
8
2
9
1
3
6
4
7
5
3
4
7
6
9
5
8
1
2
6
3
8
2
5
9
7
4
1
4
5
6
8
1
3
2
7
9
9
1
2
4
7
6
3
5
8
2
9
3
5
7
4
1
5
8
1
8
4
3
2
9
6
5
7
7
6
5
9
8
1
2
4
3
deepseek/deepseek-chat-v3-0324
deepseek
5
6
1
7
8
5
4
3
2
8
2
9
1
3
4
6
7
5
3
4
7
6
2
9
8
1
2
9
3
8
2
6
5
7
4
1
4
7
6
9
1
8
5
2
3
5
1
2
4
6
7
9
8
6
2
9
3
5
7
6
1
5
8
6
8
4
3
9
2
7
6
4
7
5
6
8
4
1
2
9
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.