Released: 2026-04-12
google/gemini-3-flash-preview
9
1
4
7
5
3
8
6
2
8
2
3
9
1
6
5
4
7
5
7
6
8
2
4
1
9
3
7
9
2
1
8
5
4
3
6
1
3
5
2
4
9
6
7
8
4
6
8
6
3
7
2
1
9
6
4
9
3
7
2
5
8
1
2
8
7
6
9
1
2
5
4
3
5
1
4
6
8
9
2
7
mistralai/mistral-medium-3
mistral
9
5
4
7
2
3
8
6
1
2
8
3
9
1
6
5
4
7
7
6
1
5
4
8
2
9
3
3
7
2
1
8
9
4
6
5
1
9
8
2
5
4
6
3
7
4
6
5
3
7
2
9
1
8
6
4
9
3
8
2
1
7
5
5
2
7
8
9
1
2
5
4
3
1
8
4
6
5
7
2
9
google/gemini-2.0-flash-001
9
1
4
7
2
3
5
6
8
8
6
3
5
1
9
4
4
7
2
7
5
8
4
6
1
9
3
5
9
2
6
8
1
7
3
4
1
3
8
9
7
4
6
2
5
4
7
6
2
3
5
8
1
9
6
4
9
3
5
2
1
7
8
7
8
7
1
9
4
2
5
6
3
2
1
4
6
8
9
3
2
anthropic/claude-3.7-sonnet
anthropic
9
6
1
7
4
3
5
8
2
8
2
3
5
1
9
6
4
7
7
5
4
6
2
8
1
9
3
5
9
2
1
8
7
3
6
4
1
7
8
9
3
4
6
2
5
4
3
6
2
5
6
8
1
9
6
4
9
3
7
2
5
8
1
2
8
7
4
9
1
2
5
6
3
1
5
8
6
4
9
7
4
openai/gpt-4o
openai
9
1
8
7
2
3
4
6
5
2
5
3
6
1
9
8
4
7
7
6
4
5
8
1
3
9
2
5
3
2
9
8
1
7
2
4
1
7
9
4
3
5
6
8
2
4
8
6
2
7
6
5
1
9
6
4
1
3
5
2
7
2
8
8
2
7
1
9
4
2
5
3
3
9
5
8
6
7
1
2
6
meta-llama/llama-3.3-70b-instruct
meta
9
8
7
7
5
3
1
2
4
2
5
3
9
1
6
8
4
7
4
1
6
2
8
5
3
9
7
8
3
2
1
4
9
5
6
7
1
9
5
6
7
8
6
3
2
4
7
1
3
2
5
9
1
9
6
4
9
3
1
2
7
8
5
5
6
8
4
9
7
2
1
3
3
2
4
5
6
1
7
9
8
deepseek/deepseek-chat-v3-0324
deepseek
9
6
4
7
2
3
8
5
1
2
8
3
6
1
9
5
4
7
5
7
1
8
4
6
3
9
2
7
3
2
9
8
1
4
6
5
1
9
5
2
3
4
6
8
7
4
8
6
5
7
8
3
1
9
6
4
8
3
5
2
1
7
9
8
1
7
4
9
6
2
5
3
3
2
9
1
6
7
4
8
5
x-ai/grok-3-mini-beta
x-ai
9
5
4
7
6
3
1
2
8
2
6
3
8
1
5
9
4
7
8
7
1
4
2
9
5
3
6
5
3
2
1
8
4
7
6
9
1
4
8
2
3
7
6
9
5
4
9
7
5
6
1
3
8
1
6
4
9
3
5
2
8
7
1
8
1
7
4
9
6
2
5
3
3
2
5
6
7
8
4
1
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.