Released: 2026-07-15
google/gemini-3-flash-preview
5
7
8
4
1
3
6
2
9
4
2
6
8
9
5
3
1
7
3
9
1
2
7
6
8
5
4
8
5
2
3
4
7
9
6
1
6
4
3
5
9
1
2
8
7
1
1
9
6
2
8
5
4
3
2
8
5
1
6
9
4
7
3
9
1
7
7
3
4
5
2
6
7
6
4
7
8
2
1
9
5
openai/gpt-4o
openai
5
7
8
3
2
6
4
1
9
4
2
6
8
9
5
3
7
1
3
9
1
2
7
4
8
5
6
6
5
7
1
4
8
9
3
2
9
1
3
5
6
1
2
8
4
8
4
2
6
3
9
7
1
5
2
8
5
4
6
3
1
9
7
1
6
4
7
5
2
9
2
8
7
3
9
9
8
7
5
4
3
mistralai/mistral-medium-3
mistral
6
7
8
1
3
4
5
2
9
4
1
2
8
9
5
3
6
7
3
9
5
2
7
6
4
5
8
7
2
6
3
4
8
9
1
5
9
4
3
5
6
1
2
8
7
8
5
1
6
2
9
7
3
4
2
3
5
9
6
7
8
4
1
1
6
4
7
8
3
5
9
2
5
8
9
4
1
2
6
7
3
meta-llama/llama-3.3-70b-instruct
meta
5
7
8
3
1
6
4
2
9
4
3
2
8
9
5
3
1
7
1
9
6
2
7
4
8
5
3
8
1
4
9
4
3
9
6
2
3
5
3
5
2
1
2
8
6
9
6
1
6
5
8
7
4
5
2
4
5
1
6
9
3
7
1
6
8
9
4
3
2
1
5
8
7
2
3
5
8
7
6
9
4
deepseek/deepseek-chat-v3-0324
deepseek
5
7
8
1
3
6
4
2
9
4
2
6
8
9
5
3
1
7
3
9
1
2
7
4
6
5
8
8
1
2
7
4
3
9
6
5
6
5
3
9
2
1
7
8
4
9
4
7
6
5
8
2
3
1
2
8
5
3
6
9
1
7
4
1
3
4
5
1
7
8
4
6
7
6
9
4
8
2
1
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.