Released: 2026-09-18
mistralai/mistral-medium-3
mistral
5
6
2
4
3
9
1
8
7
1
7
9
8
5
2
3
6
4
3
4
8
6
1
7
2
5
9
6
8
5
1
4
3
7
9
2
2
3
1
7
9
5
4
2
6
4
9
7
2
6
8
5
3
1
7
5
3
9
8
4
8
1
2
8
2
6
3
2
1
9
4
5
9
1
4
5
7
6
3
2
8
openai/gpt-4o
openai
3
6
5
2
7
4
1
8
9
9
4
1
8
5
3
2
6
7
2
7
8
6
1
9
3
5
4
6
8
7
1
9
2
4
3
5
5
3
9
7
4
6
6
2
1
4
1
2
3
6
5
7
9
8
7
5
6
4
3
1
8
7
2
1
9
3
5
2
8
9
4
6
8
2
4
9
6
7
5
1
3
deepseek/deepseek-chat-v3-0324
deepseek
2
6
4
3
5
9
1
8
7
5
9
3
4
7
8
2
6
1
7
1
8
6
2
1
3
5
4
6
8
2
1
9
4
5
3
9
9
3
1
7
8
5
4
2
6
4
7
5
2
6
3
8
7
9
1
5
7
8
3
6
9
4
2
3
4
6
5
2
1
7
9
8
8
2
9
4
1
7
6
3
5
meta-llama/llama-3.3-70b-instruct
meta
5
6
7
4
3
2
1
8
9
2
8
9
1
5
6
4
7
3
3
4
8
6
9
7
2
5
1
6
1
5
7
8
4
3
9
2
9
3
2
7
1
5
8
6
4
4
7
1
9
6
3
5
2
8
8
5
6
2
4
1
9
3
7
1
9
4
3
7
8
6
5
2
7
2
3
5
2
9
4
1
6
google/gemini-3-flash-preview
5
6
2
3
7
4
1
8
9
1
4
7
5
9
8
2
6
3
3
9
8
6
1
2
7
5
4
6
7
5
1
8
9
4
3
2
8
3
1
7
4
5
6
2
9
4
2
9
2
6
3
5
7
1
9
5
6
4
3
1
8
7
2
7
1
3
8
2
6
9
4
5
2
8
4
9
5
7
5
1
6
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.