Released: 2026-07-17
google/gemini-3-flash-preview
7
8
4
2
3
5
6
9
1
1
5
2
7
4
9
3
8
6
9
6
3
8
5
1
2
4
7
8
4
9
6
2
3
5
7
1
6
3
7
8
9
4
1
2
5
5
2
1
5
1
7
8
3
4
2
7
5
3
1
8
9
6
4
3
9
5
1
6
2
4
7
8
4
1
8
9
7
6
5
3
2
openai/gpt-4o
openai
7
8
2
4
3
5
6
9
1
1
5
6
7
2
9
3
8
4
9
3
4
6
8
1
2
5
7
8
4
9
1
7
2
5
3
6
6
3
7
8
9
5
1
2
4
5
2
1
9
6
7
8
4
3
2
7
5
3
4
8
9
6
1
3
9
1
5
6
4
7
1
8
4
6
8
2
7
3
1
5
9
mistralai/mistral-medium-3
mistral
2
3
4
5
8
6
9
1
7
6
7
9
1
2
3
4
5
8
8
5
1
4
9
7
2
3
6
7
2
3
6
4
5
8
9
1
4
8
6
9
7
1
3
2
5
9
1
5
2
3
8
6
7
4
3
6
7
8
5
2
1
4
9
1
4
2
3
6
9
5
8
7
5
9
8
7
1
4
2
6
3
meta-llama/llama-3.3-70b-instruct
meta
5
3
4
6
7
8
6
9
1
6
7
2
1
9
5
3
8
4
9
8
1
3
4
1
2
5
7
8
4
9
5
3
6
1
7
2
3
1
6
8
9
4
5
2
8
5
2
1
4
7
7
9
3
6
1
7
5
3
2
9
4
6
8
2
9
8
7
6
3
4
1
5
4
5
3
2
8
6
7
9
1
deepseek/deepseek-chat-v3-0324
deepseek
7
8
3
2
4
5
6
9
1
6
1
2
7
3
9
5
8
4
9
5
4
6
8
1
2
3
7
8
4
9
1
2
6
3
7
5
3
6
7
8
9
4
1
2
9
5
2
1
9
5
7
8
4
6
2
7
5
3
1
8
9
6
4
1
9
6
5
6
2
4
1
3
4
3
8
4
7
3
7
5
2
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.