Released: 2026-06-12
google/gemini-3-flash-preview
1
3
9
5
6
2
7
8
4
5
2
4
8
7
1
3
6
9
3
6
8
9
4
3
5
1
2
6
8
7
3
2
9
4
5
1
4
1
5
6
3
7
8
2
9
9
2
3
1
5
4
6
7
3
2
5
1
4
8
3
9
7
6
8
4
2
7
9
6
1
3
5
7
9
6
2
1
5
4
3
8
openai/gpt-4o
openai
1
2
9
5
6
3
4
8
7
5
7
4
8
1
2
3
6
9
3
6
8
4
7
9
5
1
2
6
3
7
9
2
8
1
5
4
4
1
5
6
3
7
2
9
8
9
8
2
1
5
4
6
7
3
2
4
1
3
8
5
7
2
6
8
5
2
7
9
6
9
3
1
7
9
6
2
4
1
8
2
5
mistralai/mistral-medium-3
mistral
1
2
9
3
6
4
7
8
5
5
4
6
8
7
9
3
1
2
3
6
8
5
1
7
5
9
2
5
8
7
6
2
3
1
4
9
4
1
3
2
9
8
6
5
7
9
3
2
1
5
4
8
7
3
2
5
4
3
8
1
9
6
7
6
9
2
7
9
5
4
3
8
7
8
1
4
3
6
2
5
9
deepseek/deepseek-chat-v3-0324
deepseek
1
2
9
3
6
5
4
8
7
5
7
4
2
8
9
3
6
1
3
6
8
4
7
1
5
9
2
6
3
7
8
2
4
9
1
5
4
1
5
9
3
7
6
2
8
9
8
2
1
5
6
4
7
3
2
5
1
6
8
3
7
4
9
8
4
2
7
9
6
1
5
3
7
9
6
5
4
2
8
3
1
meta-llama/llama-3.3-70b-instruct
meta
1
5
9
8
6
4
2
3
7
8
3
6
9
1
7
5
4
2
4
6
8
3
5
2
1
9
7
3
9
7
4
2
8
6
1
5
5
1
2
6
9
3
8
7
4
2
8
4
1
7
5
9
6
3
9
7
1
5
8
6
4
2
3
6
4
3
7
5
1
2
8
9
7
2
5
4
3
9
1
6
8
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.