Released: 2026-07-10
google/gemini-3-flash-preview
3
4
8
6
1
2
7
9
5
7
9
1
5
4
2
8
6
3
6
2
5
8
9
3
4
2
1
1
8
6
3
5
4
9
2
7
2
3
4
7
2
9
5
8
6
5
7
9
1
8
6
3
4
2
8
6
3
9
7
5
2
1
4
7
5
9
2
3
1
6
8
4
4
1
2
4
6
8
5
7
9
openai/gpt-4o
openai
3
4
1
8
2
7
6
9
5
7
9
5
1
6
4
8
3
2
6
2
8
5
9
3
7
4
1
1
8
7
6
5
9
4
2
3
5
3
4
7
1
2
9
8
6
2
6
9
4
8
6
3
5
7
8
7
3
9
4
5
1
6
2
4
5
6
2
7
1
5
1
9
9
1
2
3
3
8
5
7
4
mistralai/mistral-medium-3
mistral
3
4
7
8
2
9
6
1
5
5
9
1
6
7
4
8
3
2
6
2
8
5
4
3
7
9
1
1
3
8
4
5
7
9
2
6
2
7
4
1
9
8
5
6
3
9
6
5
3
8
2
4
1
7
8
1
3
9
6
5
2
4
7
4
5
2
7
3
1
6
8
9
7
8
9
2
1
6
3
5
4
meta-llama/llama-3.3-70b-instruct
meta
5
4
8
6
3
2
1
9
7
9
1
6
4
7
8
5
3
2
6
2
7
1
9
5
4
8
3
1
8
5
3
4
6
2
7
9
3
9
4
2
8
1
7
6
5
2
7
1
5
6
9
8
4
3
8
5
3
9
1
7
6
2
4
4
6
2
8
5
3
9
1
7
7
3
9
4
2
1
5
8
6
deepseek/deepseek-chat-v3-0324
deepseek
3
4
8
1
6
9
2
7
5
5
9
1
7
2
4
8
6
3
6
2
7
8
5
3
4
9
1
1
3
6
4
5
7
9
2
8
2
7
4
3
9
8
5
1
6
9
8
5
2
1
6
3
4
7
8
1
3
9
7
5
6
4
2
7
5
9
6
4
2
1
3
8
4
6
2
5
3
8
9
7
1
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.