Released: 2026-03-29
google/gemini-3-flash-preview
1
2
5
7
8
3
6
9
4
6
3
7
4
9
2
8
5
1
8
4
9
5
6
1
7
3
2
2
8
4
1
3
9
5
7
6
7
1
6
8
5
2
9
4
3
5
9
3
6
4
8
2
1
7
3
7
5
2
1
6
4
8
9
9
5
8
7
2
4
1
6
3
4
6
2
3
7
5
3
2
8
google/gemini-2.0-flash-001
2
9
5
7
6
3
1
8
4
6
3
1
4
8
9
8
5
7
8
4
7
5
2
1
3
6
9
5
8
9
1
3
2
4
7
6
7
2
6
3
5
4
9
1
8
4
1
3
9
6
8
2
5
7
3
6
4
2
9
5
7
8
1
9
5
8
6
1
7
4
9
3
8
1
2
3
7
6
5
4
9
anthropic/claude-3.7-sonnet
anthropic
5
8
1
7
2
9
6
3
4
6
9
7
3
4
5
8
2
1
3
4
2
5
8
1
7
6
9
9
5
4
1
3
7
5
8
2
7
2
6
8
5
4
9
1
3
1
3
3
9
6
8
2
4
7
4
6
5
2
9
3
1
7
8
2
7
8
6
1
4
3
9
3
8
1
9
4
7
6
5
3
2
mistralai/mistral-medium-3
mistral
8
3
2
7
9
6
1
5
4
6
7
9
4
1
5
8
3
2
5
4
1
5
8
1
7
6
9
4
9
5
1
3
7
6
2
8
7
2
6
8
5
3
9
4
1
1
8
3
9
6
4
2
7
5
9
5
4
2
8
6
3
1
7
2
6
8
3
4
9
5
7
1
3
1
7
5
2
8
4
9
6
openai/gpt-4o
openai
3
8
1
7
2
9
6
5
4
6
5
7
4
3
2
8
1
9
9
4
2
5
8
1
3
7
6
2
3
9
1
4
6
7
8
5
7
8
6
3
5
7
9
4
2
1
9
3
7
6
8
2
4
7
4
6
5
2
9
3
1
2
8
5
7
8
6
1
4
4
9
3
8
1
4
9
7
5
5
6
7
deepseek/deepseek-chat-v3-0324
deepseek
1
5
2
7
8
6
3
9
4
6
3
9
4
2
7
8
5
1
8
4
7
5
9
1
6
2
3
5
9
4
1
6
2
7
3
8
7
8
6
3
5
9
4
1
2
2
1
3
9
4
8
5
6
7
3
7
5
2
1
9
6
8
4
4
6
8
6
3
4
1
9
5
9
1
2
8
7
5
3
4
6
meta-llama/llama-3.3-70b-instruct
meta
5
8
9
7
2
6
1
3
4
6
3
2
9
8
5
8
7
1
4
1
7
5
3
9
2
6
8
2
9
1
1
4
3
7
5
6
7
5
6
8
5
2
9
4
3
3
4
8
6
1
7
5
2
9
1
6
5
2
9
4
3
8
7
9
7
3
4
6
1
8
1
5
8
2
4
3
7
5
6
9
1
x-ai/grok-3-mini-beta
x-ai
5
8
1
7
3
9
6
2
4
6
9
2
4
1
5
8
3
7
7
4
3
5
8
1
2
6
9
4
2
9
1
6
3
7
5
8
7
1
6
3
5
4
9
8
2
8
7
3
9
4
6
2
1
5
9
5
4
2
7
8
1
3
6
2
6
8
3
1
4
5
7
9
3
1
7
8
4
2
6
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.