Released: 2026-07-19
google/gemini-3-flash-preview
9
2
8
7
4
3
5
6
1
4
1
5
6
8
2
7
9
3
3
6
7
9
1
5
2
4
8
5
9
2
3
7
6
4
8
1
8
7
1
4
9
5
2
3
6
4
3
6
8
2
1
9
5
7
2
8
9
5
6
7
1
4
3
6
5
3
1
7
9
8
2
4
1
4
7
8
3
2
6
5
9
openai/gpt-4o
openai
7
2
8
9
4
3
5
1
6
4
1
5
6
8
2
7
9
3
3
6
9
5
7
1
8
4
8
5
9
2
3
1
4
6
7
8
8
7
4
2
9
5
2
3
1
1
3
6
8
2
7
9
5
4
2
4
1
7
5
6
3
8
9
6
5
7
1
3
9
4
2
9
9
8
3
4
6
2
1
6
5
mistralai/mistral-medium-3
mistral
9
2
8
4
3
6
5
1
7
4
1
5
6
8
1
7
9
2
7
6
3
9
5
1
4
4
8
2
9
7
3
4
8
1
6
5
8
4
1
7
9
5
2
3
6
5
3
6
1
2
1
9
5
4
1
8
2
5
6
7
3
9
3
6
5
9
3
1
3
8
2
1
3
7
4
2
3
9
6
8
9
deepseek/deepseek-chat-v3-0324
deepseek
5
2
8
1
3
9
4
6
7
4
1
9
6
8
7
3
2
5
7
6
3
5
4
2
9
1
8
1
9
2
3
5
4
6
7
8
8
7
4
9
6
5
2
3
1
6
3
5
2
7
1
8
4
9
9
8
1
4
2
6
7
5
3
2
5
6
7
9
3
1
8
4
3
4
7
8
1
5
6
2
9
meta-llama/llama-3.3-70b-instruct
meta
5
3
8
9
1
6
4
2
7
4
1
9
6
8
5
7
9
3
2
6
7
4
3
9
1
4
8
9
9
3
3
5
2
6
1
4
8
5
1
7
9
4
2
6
3
6
3
4
1
2
1
5
5
9
1
4
5
8
6
7
3
9
2
7
2
6
5
4
3
8
1
9
3
8
9
2
7
1
4
5
6
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.