Released: 2026-02-07
mistralai/mistral-medium-3
mistral
5
1
9
7
4
8
6
3
2
6
7
2
9
8
3
4
1
5
3
4
8
6
2
5
1
9
7
1
9
2
8
3
7
5
6
4
7
3
6
2
5
4
8
9
1
4
5
8
3
1
9
7
2
6
9
2
1
4
7
6
3
8
5
5
6
4
1
9
2
8
7
3
8
7
3
5
6
1
9
4
2
google/gemini-2.0-flash-001
2
4
5
7
9
1
6
3
8
6
7
1
5
8
3
4
9
2
3
9
8
6
4
2
1
5
7
4
8
2
9
3
7
5
6
1
7
1
6
8
5
4
2
3
9
9
5
3
2
1
6
8
7
4
1
3
9
4
7
5
2
8
6
5
6
4
3
2
8
7
1
9
8
2
7
1
6
9
3
4
5
openai/gpt-4o
openai
2
1
5
7
4
6
9
3
8
6
4
7
5
9
8
2
1
3
3
9
8
2
6
1
4
5
7
7
8
2
9
3
4
5
6
1
4
3
6
8
5
2
7
9
9
1
5
9
6
1
7
8
4
2
9
2
3
4
7
5
6
8
4
5
6
4
1
8
3
9
2
7
8
7
1
3
2
9
3
7
5
anthropic/claude-3.7-sonnet
anthropic
1
4
9
7
2
6
5
8
4
6
8
5
9
4
1
2
7
3
3
7
8
6
9
4
1
5
2
4
1
2
8
3
7
9
6
5
7
3
6
2
5
9
4
1
9
8
5
9
3
1
4
8
2
6
9
2
1
4
7
8
6
3
5
5
6
4
1
8
3
7
9
2
2
9
7
5
6
2
3
4
1
meta-llama/llama-3.3-70b-instruct
meta
5
8
1
7
4
9
6
2
3
6
3
9
5
8
2
4
1
7
3
2
4
1
6
5
8
9
7
1
4
2
3
9
7
5
6
8
8
9
6
2
5
1
3
7
4
7
5
3
9
1
6
2
8
4
9
1
7
4
3
8
5
6
2
5
6
8
6
2
4
1
3
9
2
7
5
8
6
3
9
4
1
deepseek/deepseek-chat-v3-0324
deepseek
2
9
5
7
3
8
6
4
1
6
7
1
5
4
2
9
8
3
3
4
8
6
9
1
1
5
2
8
1
2
9
3
4
5
6
7
4
3
6
8
5
7
2
1
9
7
5
9
2
1
6
8
3
4
9
2
3
4
7
5
1
8
6
5
6
4
1
8
3
7
2
9
1
8
7
6
2
9
3
4
5
x-ai/grok-3-mini-beta
x-ai
1
5
4
7
9
2
6
3
8
6
7
9
1
8
5
4
2
3
3
8
2
6
4
9
1
7
5
4
1
5
8
3
7
2
9
6
7
2
6
4
5
1
8
3
9
9
3
7
5
1
6
8
4
2
5
4
1
3
2
8
7
6
9
8
6
3
2
7
9
5
1
4
2
1
7
4
6
3
9
5
8
qwen/qwen3-235b-a22b
qwen
1
2
8
7
5
3
6
9
4
6
9
5
1
2
4
7
8
3
3
7
4
6
8
9
1
5
2
8
1
2
9
3
7
5
6
4
4
8
6
2
5
7
3
1
9
7
5
9
6
1
2
8
4
3
9
3
1
4
7
5
2
8
6
5
6
4
3
9
8
2
7
1
2
4
7
5
6
1
3
9
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.