Released: 2026-07-11
google/gemini-3-flash-preview
3
9
7
8
2
4
5
1
6
6
8
5
7
9
1
4
2
3
4
2
1
6
5
3
7
9
8
5
3
9
1
4
8
2
6
7
2
1
6
3
7
5
8
4
9
7
4
8
2
6
9
3
5
1
9
5
3
5
1
7
4
8
2
1
7
4
9
8
2
6
3
5
8
6
2
4
3
6
1
7
9
openai/gpt-4o
openai
4
3
7
8
9
2
5
1
6
9
8
5
7
6
1
2
4
3
1
2
6
4
5
3
9
7
8
5
3
9
1
7
8
2
6
4
2
1
4
3
4
9
8
5
7
7
6
8
2
3
5
6
9
1
8
5
3
6
2
7
4
3
9
6
7
2
9
1
4
3
8
5
3
4
1
5
8
6
7
2
9
mistralai/mistral-medium-3
mistral
3
4
7
8
9
2
5
1
6
6
8
5
7
4
1
9
2
3
9
2
1
6
5
3
7
4
8
5
3
9
4
7
8
2
6
1
2
1
6
3
2
5
8
3
4
4
7
8
2
6
9
3
5
1
1
5
3
1
2
7
4
8
9
8
7
2
9
3
4
6
1
5
7
9
4
5
8
6
1
3
2
meta-llama/llama-3.3-70b-instruct
meta
3
8
5
1
9
6
5
4
6
1
8
9
7
4
1
3
5
3
9
2
4
8
5
3
1
7
8
5
3
9
6
2
8
2
1
4
2
1
7
5
6
4
8
3
9
4
6
3
2
1
5
9
8
7
8
5
3
4
7
7
4
2
1
7
7
1
9
8
2
6
5
3
6
4
2
3
1
9
7
9
5
deepseek/deepseek-chat-v3-0324
deepseek
4
3
7
8
9
2
5
1
6
6
8
5
7
4
1
9
2
3
9
2
1
6
5
3
7
4
8
5
3
9
1
6
8
2
7
4
2
1
6
3
7
4
8
5
9
7
4
8
2
7
9
6
3
1
1
5
3
4
2
7
4
6
9
8
7
2
9
6
5
1
3
4
3
6
4
5
1
9
3
8
7
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.