Released: 2026-07-09
google/gemini-3-flash-preview
4
6
2
1
7
8
9
3
5
8
3
5
4
6
9
2
7
1
7
9
1
2
3
5
6
8
4
2
4
7
6
5
3
1
9
8
9
1
3
8
4
7
4
6
2
5
7
8
9
1
2
3
4
6
6
8
4
3
9
1
7
5
2
3
2
9
7
8
6
5
1
4
1
5
6
5
2
4
8
3
9
openai/gpt-4o
openai
4
6
1
8
7
3
9
2
5
8
3
2
9
6
5
7
4
1
7
9
5
2
1
4
6
3
4
2
4
7
6
5
9
1
8
3
9
1
3
4
8
7
4
5
6
5
7
8
1
3
2
2
9
7
6
8
4
3
2
1
5
7
9
3
2
9
7
4
6
8
1
2
1
5
6
5
9
4
3
6
9
meta-llama/llama-3.3-70b-instruct
meta
4
2
1
3
7
9
8
6
5
8
5
6
2
6
1
4
3
7
3
9
7
8
5
4
6
2
1
2
1
4
6
3
8
5
9
7
9
7
3
5
2
6
1
8
4
5
6
8
9
1
2
7
4
3
1
4
5
7
9
3
2
1
8
6
3
2
4
8
5
9
7
1
7
8
9
1
4
7
3
5
6
mistralai/mistral-medium-3
mistral
4
1
6
8
7
9
2
3
5
8
2
7
1
6
5
3
9
4
3
9
5
2
4
8
6
7
1
2
4
9
6
3
7
1
5
8
9
6
3
5
1
8
4
2
7
5
7
8
4
9
2
3
6
1
6
8
2
3
5
1
7
4
9
7
3
4
7
2
6
5
1
8
1
5
1
9
8
4
2
7
6
deepseek/deepseek-chat-v3-0324
deepseek
4
6
1
2
7
9
3
8
5
8
3
2
4
6
5
9
7
1
7
9
5
8
3
1
6
2
4
2
4
7
6
5
3
1
9
8
9
1
3
7
8
6
4
5
2
5
6
8
1
9
2
7
4
3
6
8
9
3
1
7
5
4
2
3
2
4
7
5
8
2
1
6
1
5
7
9
2
4
8
3
9
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.