Puzzle Solving Model Ranking

See which AI models perform best on Puzzle Solving, which ones stay reliable, and where the biggest gaps appear. Sort by: Metric ↑.

Models Shown

Average Puzzle Solving Score

6.9

Best Model

Step 3.5 Flash 0.0

Failure Reasons

With failure reason Wrong answer275 With failure reason Did not follow instructions128 With failure reason API error15 With failure reason Extra formatting14 With failure reason No answer11 With failure reason Timed out11

311/311

Rank	Model	Company	Puzzle Solving Score	Score	Total Cost	Tests Correct	Response Time (avg)
#47	GPT-5.2 medium	OpenAI	7.5	8.4	$1.182	2/3	5.80s
Total Tests 3 Wrong Tests 1 Total Cost $1.182 Response Time (avg) 5.80s
#48	Qwen3.8 27B high	Qwen	7.6	8.4	~$0.287	2/3	199.9s
Total Tests 3 Wrong Tests 1 Total Cost ~$0.287 Response Time (avg) 199.9s
#52	Claude Fable 5.1 medium	Anthropic	7.6	8.3	$2.909	2/3	5.70s
Total Tests 3 Wrong Tests 1 Total Cost $2.909 Response Time (avg) 5.70s
#86	GPT-5.6 Luna high	OpenAI	7.6	7.6	$0.179	2/3	14.6s
Total Tests 3 Wrong Tests 1 Total Cost $0.179 Response Time (avg) 14.6s
#112	Gemini 3.1 Flash Lite medium	Google	7.6	7.3	$0.120	2/3	1.95s
Total Tests 3 Wrong Tests 1 Total Cost $0.120 Response Time (avg) 1.95s
#157	Gemini 3.5 Flash Lite low	Google	7.6	6.6	$0.138	2/3	1.19s
Total Tests 3 Wrong Tests 1 Total Cost $0.138 Response Time (avg) 1.19s
#180	GPT-5.6 Luna low	OpenAI	7.6	6.2	$0.051	2/3	3.59s
Total Tests 3 Wrong Tests 1 Total Cost $0.051 Response Time (avg) 3.59s
#262	DeepSeek V3.2 none	DeepSeek	7.6	5.0	$0.054	2/3	6.91s
Total Tests 3 Wrong Tests 1 Total Cost $0.054 Response Time (avg) 6.91s
#21	Gemini 3.5 Flash medium	Google	7.7	9.0	$0.665	2/3	2.38s
Total Tests 3 Wrong Tests 1 Total Cost $0.665 Response Time (avg) 2.38s
#39	Claude Fable 5 medium	Anthropic	7.7	8.6	$3.004	2/3	5.18s
Total Tests 3 Wrong Tests 1 Total Cost $3.004 Response Time (avg) 5.18s
#53	Gemini 2.5 Flash medium	Google	7.7	8.2	$0.644	2/3	3.18s
Total Tests 3 Wrong Tests 1 Total Cost $0.644 Response Time (avg) 3.18s
#55	Claude Sonnet 5 medium	Anthropic	7.7	8.2	$0.922	2/3	2.98s
Total Tests 3 Wrong Tests 1 Total Cost $0.922 Response Time (avg) 2.98s
#65	GPT-5.6 Terra high	OpenAI	7.7	8.0	$0.836	2/3	5.45s
Total Tests 3 Wrong Tests 1 Total Cost $0.836 Response Time (avg) 5.45s
#67	GPT-5.2 Chat none	OpenAI	7.7	7.9	$0.594	2/3	4.10s
Total Tests 3 Wrong Tests 1 Total Cost $0.594 Response Time (avg) 4.10s
#69	Qwen3.8 27B low	Qwen	7.7	7.9	~$0.071	2/3	4.91s
Total Tests 3 Wrong Tests 1 Total Cost ~$0.071 Response Time (avg) 4.91s

Puzzle Solving Ranking

Filter models

Top Models by Puzzle Solving Score

Puzzle Solving Score vs Total Cost

Top Models by Response Time (avg)

Puzzle Solving Ranking

Filter models

Top Models by Puzzle Solving Score

Puzzle Solving Score vs Total CostEstimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 2.8999 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.286715.

Top Models by Response Time (avg)

Puzzle Solving Score vs Total Cost