Released: 2026-07-28
google/gemini-3-flash-preview
1
4
5
6
9
2
8
3
7
2
6
8
3
5
7
9
1
4
3
7
9
1
4
8
2
6
5
9
8
2
4
2
7
3
5
1
7
3
1
8
2
9
5
4
6
5
2
4
5
6
1
8
9
3
4
9
2
5
7
6
1
3
8
8
5
7
1
3
4
6
2
9
9
1
6
2
8
3
7
5
4
openai/gpt-4o
openai
7
3
5
6
1
4
9
8
2
2
6
8
3
5
9
1
7
4
1
4
9
2
7
8
6
5
3
9
8
6
4
2
7
3
5
1
5
7
3
8
1
9
2
4
6
6
2
4
5
3
1
8
9
7
3
9
1
7
4
6
5
2
8
8
5
7
1
9
2
4
6
3
4
1
2
9
6
3
7
8
5
meta-llama/llama-3.3-70b-instruct
meta
5
3
5
6
7
8
9
1
2
2
6
1
3
5
4
8
7
9
8
9
7
1
3
2
5
6
4
9
5
3
4
1
7
6
2
8
1
8
9
5
2
9
4
3
6
4
7
2
8
6
5
3
9
1
3
9
4
2
7
1
5
8
6
6
1
8
9
4
3
7
5
2
7
2
6
5
8
6
1
4
3
mistralai/mistral-medium-3
mistral
1
3
5
6
2
4
9
7
8
2
6
4
3
9
7
1
8
5
7
8
9
1
5
8
2
3
4
9
2
8
4
3
7
3
6
1
3
7
1
8
6
9
5
4
2
5
4
4
2
6
1
8
9
7
4
9
2
5
7
3
6
1
8
8
5
7
1
4
6
9
2
3
6
1
3
9
8
2
7
5
4
deepseek/deepseek-chat-v3-0324
deepseek
4
3
5
6
1
2
9
7
8
2
6
8
3
4
5
1
9
7
1
7
9
7
3
8
4
2
5
9
8
2
4
5
7
3
6
1
7
5
3
2
8
9
6
4
7
6
1
4
5
6
3
8
7
2
5
9
3
9
7
4
2
1
8
8
2
7
1
2
6
5
3
9
3
1
6
8
9
5
7
8
4
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.