Released: 2026-02-19
mistralai/mistral-medium-3
mistral
4
5
8
6
9
7
3
1
2
9
3
2
4
8
1
7
6
5
7
6
6
2
1
3
5
4
8
1
9
3
7
6
8
8
2
4
5
7
6
1
4
9
2
3
8
3
2
4
8
5
6
6
7
9
8
6
9
3
7
4
1
5
2
2
1
7
8
3
2
4
9
6
6
4
5
9
2
1
8
8
3
google/gemini-3-flash-preview
6
5
8
2
9
4
3
1
7
9
1
3
5
7
6
8
2
4
7
4
2
8
1
3
5
9
6
1
9
6
7
3
8
4
5
2
5
8
7
4
6
9
2
3
1
3
2
4
1
5
7
6
8
9
8
6
9
3
4
5
1
7
2
4
7
5
8
9
2
9
6
3
2
3
1
6
0
1
7
8
5
google/gemini-2.0-flash-001
2
5
8
1
9
4
3
6
7
9
4
1
5
3
6
7
2
8
7
3
6
2
1
8
5
9
4
1
9
2
7
6
3
8
4
5
5
8
7
4
2
9
1
3
6
3
6
4
1
5
8
6
7
9
8
6
9
3
4
7
2
5
1
4
1
5
8
7
2
9
3
6
6
2
3
9
5
1
4
8
7
anthropic/claude-3.7-sonnet
anthropic
2
5
8
4
9
7
3
6
1
4
3
6
5
8
1
9
7
2
7
9
1
2
1
6
5
8
4
1
8
2
7
4
9
4
5
3
5
7
3
6
2
8
2
3
8
3
4
4
1
5
3
6
2
9
8
6
9
3
7
4
1
5
6
9
1
5
8
6
2
7
4
3
6
2
7
9
3
5
4
8
1
meta-llama/llama-3.3-70b-instruct
meta
4
5
8
6
9
3
3
1
2
2
3
6
4
8
5
1
9
7
7
1
5
2
1
9
5
6
4
1
9
3
7
4
2
8
5
6
5
8
4
1
3
6
2
3
9
3
2
4
5
5
1
6
8
9
9
6
9
3
7
8
4
2
1
8
4
1
8
6
2
7
5
3
6
7
2
9
5
4
1
8
5
openai/gpt-4o
openai
4
5
8
6
9
7
3
2
1
9
1
6
5
2
3
8
7
4
7
3
2
4
1
8
5
9
6
1
9
5
7
3
6
4
8
2
5
4
7
9
8
1
2
3
6
3
8
4
2
5
9
6
1
9
8
6
9
3
4
5
1
6
7
2
7
1
8
6
2
9
5
3
6
2
3
1
7
4
7
8
5
deepseek/deepseek-chat-v3-0324
deepseek
4
5
8
2
9
3
3
6
7
2
3
6
5
7
8
4
9
1
7
9
2
6
1
4
5
3
8
1
2
6
7
8
9
8
4
5
5
8
7
1
6
4
2
3
9
3
7
4
8
5
2
6
1
9
8
6
9
3
4
7
1
5
2
7
4
3
8
6
2
9
7
4
9
2
5
4
3
1
7
8
6
x-ai/grok-3-mini-beta
x-ai
4
5
8
6
9
1
3
7
2
6
1
2
4
3
7
9
5
8
7
9
3
2
1
8
5
4
6
1
8
6
7
4
5
8
2
3
5
7
4
9
8
6
2
3
1
3
2
1
8
5
4
6
7
9
2
4
7
3
6
9
1
8
5
9
3
5
1
7
2
4
6
8
8
6
9
5
2
3
7
1
4
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.