Released: 2026-06-20
google/gemini-3-flash-preview
6
4
5
7
2
9
8
3
1
2
9
8
5
6
1
3
7
4
7
3
1
8
3
4
5
6
2
4
6
2
3
8
5
9
1
7
8
7
3
6
9
2
4
1
5
1
5
9
1
4
7
6
2
8
3
2
4
9
1
8
1
5
6
9
1
7
4
5
6
2
8
3
5
8
6
2
7
3
1
9
4
openai/gpt-4o
openai
1
6
5
7
2
9
4
3
8
2
4
8
1
6
5
3
7
9
9
3
7
8
4
3
5
6
2
6
1
4
3
8
7
9
2
5
8
7
2
6
9
3
7
1
5
3
5
9
2
4
1
6
8
7
4
2
3
9
3
8
1
5
6
7
9
1
4
5
6
2
8
3
5
8
6
5
7
2
7
9
1
meta-llama/llama-3.3-70b-instruct
meta
5
1
9
7
3
9
4
8
6
3
4
8
2
1
5
3
7
9
6
3
2
9
5
8
1
6
2
9
2
7
3
8
1
5
4
6
1
6
5
6
9
4
8
1
5
4
5
3
1
4
6
2
9
8
8
9
1
5
2
7
6
3
4
2
7
6
4
5
3
9
1
8
7
8
4
8
6
2
5
9
1
mistralai/mistral-medium-3
mistral
6
1
5
7
4
9
8
3
2
7
2
8
1
6
5
3
4
9
9
3
4
8
2
3
7
6
2
2
6
7
3
8
1
4
5
9
8
4
3
6
9
2
7
1
5
1
5
9
4
4
7
2
8
6
3
7
2
9
6
8
1
5
4
4
9
1
2
5
6
3
7
3
5
8
6
3
7
4
9
9
1
deepseek/deepseek-chat-v3-0324
deepseek
4
2
5
7
6
9
8
3
1
6
7
8
5
1
2
3
4
9
9
3
1
8
3
4
7
6
2
1
6
2
3
8
5
9
7
4
7
4
3
6
9
1
2
1
5
8
5
9
2
4
7
6
8
3
3
9
7
4
2
8
1
5
6
2
1
4
9
5
6
7
2
3
5
8
6
1
7
3
4
9
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.