Released: 2026-09-09
mistralai/mistral-medium-3
mistral
9
5
7
3
6
1
2
8
4
2
6
4
9
7
8
3
5
1
8
3
9
5
4
2
1
6
7
1
2
6
4
3
5
8
9
7
3
4
5
1
9
6
7
2
8
4
8
7
2
1
9
5
3
6
5
1
8
6
2
3
4
7
9
6
9
2
7
8
4
9
1
5
7
9
3
8
5
1
6
4
2
meta-llama/llama-3.3-70b-instruct
meta
5
3
9
8
6
1
2
7
4
6
8
7
3
4
2
1
5
9
2
9
1
5
8
7
4
6
3
1
5
3
9
2
6
8
4
7
3
2
6
1
5
4
9
8
7
4
7
8
6
3
5
9
1
2
9
1
4
7
8
3
6
2
5
8
6
5
2
1
9
7
3
4
7
4
2
4
9
8
5
1
6
openai/gpt-4o
openai
5
8
1
6
3
9
2
7
4
7
6
9
4
1
2
8
5
3
3
4
2
5
8
7
1
6
9
1
2
6
3
9
4
5
8
7
3
7
5
1
4
6
8
9
2
4
9
8
2
5
7
3
1
6
9
1
8
7
2
3
4
6
5
2
3
4
7
6
8
9
5
1
6
5
7
9
5
1
2
4
8
google/gemini-3-flash-preview
5
8
1
6
7
3
2
9
4
7
6
4
9
1
2
8
5
3
9
3
2
5
8
4
1
6
7
1
2
6
8
4
5
3
9
7
3
7
9
1
2
6
4
8
5
4
5
8
3
9
7
6
2
1
6
1
8
4
2
9
5
7
3
8
4
5
7
3
1
9
0
0
2
9
3
0
5
0
7
4
8
deepseek/deepseek-chat-v3-0324
deepseek
8
9
5
3
6
1
2
7
4
7
6
1
4
9
2
8
5
3
3
2
4
5
8
7
1
6
9
1
8
2
9
4
5
6
3
7
3
7
9
1
2
6
4
8
5
4
5
6
8
7
3
9
2
1
9
1
8
6
2
4
5
3
7
5
4
7
7
3
8
9
1
6
6
3
3
2
5
9
7
4
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.