Released: 2026-07-25
google/gemini-3-flash-preview
1
5
7
3
2
8
6
4
9
6
9
4
7
1
9
3
2
5
3
2
8
6
5
4
1
9
7
5
7
9
1
3
2
4
6
8
4
3
6
8
9
5
7
6
2
2
8
1
4
7
6
5
1
3
7
4
3
9
4
1
2
5
6
8
1
2
5
6
3
9
7
4
9
6
5
2
8
7
3
1
1
openai/gpt-4o
openai
1
7
6
5
2
8
3
4
9
5
9
4
7
6
3
1
2
8
8
2
3
1
5
9
6
7
7
6
5
1
4
3
7
8
9
2
4
3
7
8
9
5
5
6
1
2
8
9
3
7
6
4
1
5
7
4
8
9
1
2
5
3
6
3
1
2
6
4
3
9
8
7
9
6
5
2
8
1
7
5
4
meta-llama/llama-3.3-70b-instruct
meta
1
5
8
6
4
3
2
7
9
4
2
9
1
8
5
6
3
7
3
6
7
9
5
2
1
4
8
8
4
5
7
3
9
6
1
2
2
1
6
8
9
7
5
3
4
9
3
7
4
1
6
8
5
2
5
8
1
3
2
4
9
6
7
6
9
4
5
7
1
3
2
8
7
7
3
2
6
8
4
9
5
mistralai/mistral-medium-3
mistral
1
6
3
4
2
8
5
7
9
9
7
4
1
6
5
8
2
3
8
2
5
3
9
7
1
4
6
4
9
6
7
3
1
2
5
8
5
3
7
2
8
9
6
4
1
2
8
1
5
4
6
3
9
7
7
1
2
9
5
4
3
8
6
6
4
8
3
1
2
9
7
5
3
5
9
6
7
8
4
1
2
deepseek/deepseek-chat-v3-0324
deepseek
1
5
2
7
6
8
4
3
9
6
7
4
1
9
3
5
2
8
9
3
8
4
5
2
6
1
7
5
4
9
8
3
1
7
6
2
3
6
1
2
7
4
8
9
5
2
8
7
5
9
6
3
1
4
4
2
3
6
1
5
9
8
7
7
9
6
8
2
3
1
5
4
8
1
5
9
4
7
2
6
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.