Released: 2026-03-27
google/gemini-3-flash-preview
8
2
6
7
3
9
5
4
1
9
5
4
6
1
2
7
3
8
7
3
1
8
4
5
2
9
6
2
4
9
5
6
3
1
8
7
6
1
7
2
8
9
9
3
4
3
8
5
1
9
4
6
7
2
4
9
3
3
7
6
8
2
5
5
7
2
4
4
1
3
6
9
1
6
8
9
2
7
4
5
3
google/gemini-2.0-flash-001
8
2
6
7
4
9
5
3
1
9
5
4
3
1
2
6
7
8
1
3
7
6
5
8
2
9
4
7
4
9
5
2
3
1
8
6
2
6
1
9
8
7
5
4
3
3
8
5
1
6
4
9
2
7
4
9
3
8
7
6
2
1
5
5
7
2
4
9
1
3
6
9
6
1
8
2
3
5
4
7
9
openai/gpt-4o
openai
8
2
6
7
3
9
5
4
1
9
5
7
6
1
2
4
3
8
1
3
4
8
5
6
2
9
7
2
4
9
5
7
3
1
8
6
7
1
8
2
8
4
6
3
5
3
6
5
1
9
8
7
2
4
4
9
3
3
6
7
8
1
5
5
8
2
4
2
1
3
6
9
6
7
1
9
4
5
2
7
3
anthropic/claude-3.7-sonnet
anthropic
8
6
1
7
5
9
4
2
3
9
7
4
3
1
2
5
9
8
5
3
7
4
6
8
2
1
9
2
8
9
5
4
3
1
7
6
1
9
6
2
8
7
9
3
4
7
4
5
1
9
6
8
5
2
4
1
3
6
2
5
7
8
5
7
5
2
8
7
1
3
6
9
6
2
8
9
3
4
5
4
1
meta-llama/llama-3.3-70b-instruct
meta
8
5
6
7
4
9
1
2
3
9
4
1
5
1
2
7
8
6
5
3
7
9
6
8
2
4
1
1
9
9
3
5
6
4
7
2
6
2
4
1
8
7
5
3
9
7
8
5
2
9
4
6
1
3
4
1
3
6
2
5
9
8
7
2
6
9
8
7
1
3
5
4
3
7
8
4
5
6
2
1
9
mistralai/mistral-medium-3
mistral
8
1
4
7
5
9
6
2
3
9
6
7
3
1
2
4
5
8
2
3
5
4
6
8
9
7
1
5
2
9
8
4
3
1
6
7
1
7
6
5
8
2
9
3
4
3
4
8
9
2
6
7
1
5
4
9
3
2
7
1
8
5
6
7
8
2
6
3
5
4
9
1
6
5
1
4
9
7
2
8
3
deepseek/deepseek-chat-v3-0324
deepseek
8
2
4
7
3
9
5
1
6
9
5
6
4
1
2
7
3
8
1
3
7
8
5
6
2
4
9
2
4
9
6
7
3
1
5
8
6
1
3
5
8
9
2
7
4
7
8
5
1
2
4
6
9
3
4
6
8
9
7
2
3
8
5
5
7
2
3
4
1
8
6
9
3
9
1
2
6
8
4
5
7
x-ai/grok-3-mini-beta
x-ai
8
1
6
7
3
9
4
5
2
9
4
7
5
1
2
6
3
8
5
3
2
6
4
8
9
1
7
2
7
9
1
5
3
8
4
6
1
6
8
2
9
7
5
3
4
3
2
5
4
6
1
7
8
9
4
8
1
3
2
6
9
7
5
7
5
4
9
8
3
1
6
9
6
9
3
8
7
5
2
4
1
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.