Released: 2026-08-16
openai/gpt-4o
openai
4
5
7
8
9
3
1
2
6
3
8
1
2
6
5
7
4
9
2
6
9
1
4
7
3
8
5
9
4
8
5
1
6
2
7
3
5
2
3
4
8
9
6
1
7
6
1
7
3
2
8
5
9
4
7
9
4
6
5
2
8
3
1
8
3
5
9
7
1
4
6
2
1
2
6
7
3
4
9
5
8
google/gemini-3-flash-preview
5
7
9
1
2
8
3
4
6
6
8
1
3
4
5
7
2
9
2
4
3
9
6
7
1
8
5
9
3
8
7
1
6
2
5
4
5
2
7
4
8
3
6
9
1
4
1
6
2
9
5
8
7
3
7
9
4
6
5
2
8
3
1
8
6
5
7
3
1
9
4
2
1
5
2
8
4
9
3
6
7
mistralai/mistral-medium-3
mistral
4
3
7
8
2
1
5
6
9
6
8
1
3
5
2
4
7
9
5
2
9
6
4
7
3
8
1
9
7
8
5
1
6
2
4
3
3
2
6
4
8
9
7
1
5
1
5
4
2
7
3
8
9
6
7
9
3
6
8
4
5
3
1
8
4
5
1
3
5
9
2
7
1
6
2
7
9
8
3
5
4
meta-llama/llama-3.3-70b-instruct
meta
5
3
4
8
2
6
1
9
6
8
2
1
5
4
5
3
7
9
6
9
7
3
1
7
5
8
4
9
5
8
7
1
6
2
4
3
4
2
3
4
8
5
6
1
5
1
7
6
2
9
3
4
5
8
7
9
5
6
3
1
8
3
1
3
1
9
4
5
8
7
2
6
2
8
6
1
7
9
5
3
4
deepseek/deepseek-chat-v3-0324
deepseek
3
5
7
9
2
8
1
4
6
4
8
1
3
6
5
7
2
9
2
6
9
1
4
7
5
8
3
9
3
8
5
1
6
2
7
4
5
2
6
4
8
3
9
1
7
8
1
4
7
9
2
3
6
5
7
9
3
6
5
4
8
3
1
6
4
5
2
3
1
4
9
8
1
7
2
8
7
9
6
5
2
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.