Released: 2026-03-12
mistralai/mistral-medium-3
mistral
9
8
6
1
2
4
7
3
5
2
7
1
3
6
5
9
4
8
5
4
3
7
9
6
8
1
2
8
9
2
7
5
3
6
4
1
6
3
7
8
4
9
5
2
1
1
5
4
6
8
2
3
7
9
7
2
1
5
3
4
8
9
6
4
6
5
2
1
9
3
8
7
3
1
8
9
7
6
2
5
4
openai/gpt-4o
openai
1
8
6
9
3
5
7
4
2
2
5
7
3
6
1
9
8
8
3
4
9
8
7
6
2
1
5
8
9
2
7
5
3
6
2
4
5
3
7
2
1
9
2
7
6
6
1
4
5
8
4
8
3
9
9
6
1
5
4
2
3
9
7
4
7
5
6
1
7
8
6
3
7
2
3
1
9
8
4
5
1
google/gemini-2.0-flash-001
5
8
6
9
2
1
7
4
3
2
1
9
3
7
4
5
6
8
4
7
3
8
6
5
9
1
2
8
9
2
7
5
3
6
1
4
7
3
1
6
4
9
8
2
5
6
5
4
1
8
2
9
3
7
9
6
1
5
3
7
2
8
4
3
4
5
2
1
9
8
6
7
2
1
7
4
9
7
3
5
6
google/gemini-3-flash-preview
1
8
3
9
4
5
7
2
6
2
6
7
3
8
1
5
4
8
5
4
9
8
7
6
2
1
3
8
9
2
7
5
3
6
4
1
7
3
6
4
2
9
8
5
0
1
5
4
6
8
2
9
3
7
6
2
1
5
3
4
9
7
8
4
7
5
2
1
8
3
6
9
9
1
3
6
7
8
4
5
2
anthropic/claude-3.7-sonnet
anthropic
9
8
6
1
2
5
7
4
3
2
5
7
3
4
9
1
6
8
3
4
1
8
7
6
5
1
9
7
9
2
7
5
3
6
8
1
5
3
8
6
1
9
4
2
7
1
6
4
4
8
2
9
3
5
6
2
1
5
9
4
8
7
2
4
7
5
2
1
3
9
6
4
8
1
9
9
6
8
3
5
7
meta-llama/llama-3.3-70b-instruct
meta
5
8
6
4
3
7
7
2
9
2
4
3
3
1
5
8
9
8
1
5
9
9
8
6
4
1
3
4
9
2
7
5
1
6
8
5
7
3
5
1
4
9
2
6
8
8
1
4
6
2
3
5
3
4
9
6
1
5
7
8
3
4
2
3
2
5
2
1
4
1
6
9
6
7
8
9
6
2
9
5
1
deepseek/deepseek-chat-v3-0324
deepseek
4
8
6
9
2
1
7
3
5
2
1
7
3
4
5
6
9
8
5
3
9
8
7
6
4
1
2
1
9
2
7
5
3
6
8
4
7
3
8
6
4
9
5
2
1
6
5
4
1
8
2
9
3
7
3
6
1
5
9
7
8
4
2
9
2
5
2
1
4
3
6
7
8
7
3
4
6
8
2
5
9
x-ai/grok-3-mini-beta
x-ai
5
8
9
1
3
4
7
2
6
2
1
7
3
6
5
4
9
8
3
4
8
9
2
6
5
1
7
8
9
2
7
5
1
6
4
3
6
3
5
8
4
9
1
7
2
1
7
4
6
9
2
8
3
5
7
6
1
5
8
3
9
4
2
4
2
5
3
1
7
9
6
8
9
5
6
4
7
8
3
1
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.