Released: 2026-07-13
google/gemini-3-flash-preview
5
4
3
2
9
6
7
8
1
1
9
2
7
4
8
3
6
5
7
8
6
3
1
5
2
9
4
3
2
9
6
8
7
1
5
4
4
7
5
9
3
1
8
6
2
8
6
1
4
2
9
5
3
7
9
5
8
1
6
3
4
2
7
2
3
7
4
5
8
6
1
9
6
1
4
5
7
2
9
3
8
meta-llama/llama-3.3-70b-instruct
meta
5
4
8
6
3
2
1
9
7
1
7
3
7
9
5
3
2
5
9
5
2
1
7
8
4
6
3
8
2
9
5
6
1
7
5
4
4
1
5
3
4
9
2
6
8
6
8
7
4
2
3
5
1
9
7
9
1
8
5
6
3
4
2
2
6
4
9
1
7
8
3
5
3
3
6
2
8
4
9
7
1
openai/gpt-4o
openai
3
4
2
9
5
6
8
1
7
1
9
8
7
4
2
3
6
5
5
7
6
3
1
8
4
2
9
6
2
9
1
8
7
5
3
4
4
8
5
2
3
1
7
6
9
7
1
3
4
2
5
6
9
8
9
5
7
8
6
3
1
4
2
2
6
4
5
7
9
8
9
1
8
3
1
6
9
4
9
7
8
deepseek/deepseek-chat-v3-0324
deepseek
5
4
3
6
8
9
7
2
1
1
8
6
7
4
2
3
9
5
7
9
2
3
1
5
8
4
6
6
2
9
1
8
7
4
5
3
4
3
5
9
6
1
2
8
7
8
1
7
4
2
3
6
5
9
9
7
8
2
6
4
1
3
5
2
5
1
8
3
6
9
7
4
3
6
4
5
9
1
2
7
8
mistralai/mistral-medium-3
mistral
3
4
6
8
7
2
1
9
5
1
8
2
7
9
5
3
4
6
7
5
9
3
1
4
6
8
2
4
3
9
6
8
7
2
5
1
8
2
5
9
4
1
7
6
3
6
7
1
4
2
3
5
8
9
9
6
7
2
5
3
8
1
4
2
1
8
5
3
6
4
9
7
5
9
4
1
8
9
6
7
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.