Released: 2026-03-19
google/gemini-3-flash-preview
2
7
5
6
9
4
8
3
1
8
6
1
3
2
9
4
5
7
9
3
4
7
5
8
6
2
0
1
5
7
8
3
2
9
4
6
3
9
8
4
6
5
2
1
7
4
2
6
1
9
7
5
8
3
7
8
9
5
1
3
2
6
4
6
4
3
2
7
1
1
9
5
5
1
2
9
4
6
7
3
8
openai/gpt-4o
openai
2
9
5
7
6
4
8
3
1
8
6
1
3
2
9
4
5
7
4
3
7
1
5
8
6
2
9
1
2
9
8
3
5
7
4
6
5
7
8
4
1
6
2
9
3
3
4
6
2
9
7
5
1
8
7
8
4
5
9
3
1
6
2
6
1
3
9
7
2
8
5
4
9
5
2
6
4
1
3
7
9
mistralai/mistral-medium-3
mistral
2
4
5
6
7
9
8
3
1
8
9
1
3
2
4
6
5
7
5
3
7
1
6
8
4
2
9
1
7
3
8
5
2
9
4
6
9
5
8
4
1
6
2
7
3
4
2
6
9
3
7
5
1
8
7
8
4
5
9
3
1
6
2
6
1
9
2
8
1
3
7
5
3
6
2
9
4
5
7
8
1
anthropic/claude-3.7-sonnet
anthropic
2
4
5
7
6
9
8
3
1
9
8
1
3
2
4
6
5
7
4
3
7
1
5
8
9
2
4
1
7
3
2
9
5
4
8
6
5
9
8
4
1
6
2
7
3
4
2
6
8
3
7
5
1
9
7
1
9
5
8
3
1
6
2
6
4
4
1
7
1
3
9
5
3
5
2
9
4
2
7
4
8
google/gemini-2.0-flash-001
2
9
5
7
6
4
8
3
1
8
6
4
3
1
2
9
5
7
1
3
7
4
9
8
5
2
6
1
2
3
8
5
9
7
4
6
5
7
8
4
3
6
2
1
9
4
9
6
1
2
7
3
8
5
7
4
9
5
8
3
6
1
2
6
8
1
2
7
5
4
9
3
3
5
2
9
4
1
1
6
8
meta-llama/llama-3.3-70b-instruct
meta
2
8
5
4
6
3
7
3
1
4
1
3
3
5
9
6
5
7
3
3
1
2
8
8
5
2
9
1
9
7
5
3
6
4
8
6
8
5
6
1
9
4
2
7
3
6
4
2
7
1
5
3
9
5
7
2
9
8
4
1
1
6
5
5
6
8
9
7
2
1
4
2
9
7
4
6
2
8
8
1
6
deepseek/deepseek-chat-v3-0324
deepseek
2
8
5
7
6
4
9
3
1
4
6
1
3
2
9
8
5
7
9
3
7
1
5
8
6
2
4
1
2
4
8
9
5
3
7
6
5
7
8
4
3
6
2
1
9
3
9
6
2
1
7
5
4
8
7
4
9
5
8
3
1
6
2
6
8
3
2
7
1
4
9
5
8
5
2
9
4
6
7
1
3
x-ai/grok-3-mini-beta
x-ai
2
4
5
6
7
9
8
3
1
4
6
1
3
8
2
9
5
7
9
3
7
1
5
8
6
2
4
1
5
9
8
2
4
3
7
6
5
7
8
4
6
1
2
9
3
3
2
6
9
1
7
5
4
8
7
8
4
5
9
3
1
6
2
6
9
3
2
4
8
7
1
5
8
1
2
7
3
6
4
5
9
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.