✨ Updated on August 3rd, 2025 - Latest LLM data and pricing

Model Efficiency
Radar Analysis

Compare up to 5 models across key performance dimensions. Each metric is normalized to show relative strengths and weaknesses.

Quick Comparisons

Select pre-configured model groups for common comparison scenarios

Select Models to Compare
0/5 selected

Choose up to 5 models to visualize their efficiency profiles side-by-side

Grok 4
xAI
Q: 68$6
68 t/s10.0s
o3-pro
OpenAI
Q: 68$35
33 t/s85.8s
o3
OpenAI
Q: 67$4
200 t/s13.0s
Gemini 2.5 Pro (AI Studio)
Google
Q: 65$3
147 t/s38.0s
o4-mini (high)
OpenAI
Q: 65$2
114 t/s47.8s
Gemini 2.5 Flash (Reasoning) (AI Studio)
Google
Q: 58$1
356 t/s9.4s
DeepSeek R1 0528 (May '25)
DeepSeek
Q: 59$4
339 t/s0.7s
Qwen3 32B (Reasoning)
Alibaba
Q: 55$1
2496 t/s0.2s
DeepSeek R1 0528
DeepSeek
Q: 59$1
22 t/s3.4s
Qwen3 235B (Reasoning)
Alibaba
Q: 56$0
79 t/s0.8s
Grok 3 mini Reasoning (high)
xAI
Q: 58$0
211 t/s0.6s
Llama 4 Scout
Meta
Q: 34$0
132 t/s0.3s
Claude 4 Sonnet Thinking
Anthropic
Q: 59$6
44 t/s1.2s
Claude 4 Opus
Anthropic
Q: 48$30
22 t/s3.6s
o3-mini
OpenAI
Q: 53$2
189 t/s11.7s
MiniMax M1 80k
MiniMax
Q: 53$1
20 t/s1.5s
Phi-4
Microsoft Azure
Q: 32$0
41 t/s0.4s
Magistral Small
Mistral
Q: 36$1
197 t/s0.3s
Gemini 2.5 Flash (AI Studio)
Google
Q: 47$0
284 t/s0.3s
QwQ-32B
Alibaba
Q: 50$0
48 t/s0.3s

Metric Definitions

Quality

Artificial Analysis Intelligence Index score. Higher values indicate better reasoning and task performance.

Speed

Output tokens per second during generation. Uses logarithmic scaling due to extreme variations (20-2500+ t/s).

Value

Inverted price score - lower cost models score higher. Represents cost-effectiveness.

Responsiveness

Inverted latency score - lower time-to-first-token scores higher. Indicates how quickly models start responding.

Context

Maximum context window size. Uses logarithmic scaling due to extreme variations (16k-10M tokens).

📊

Ready to Compare Models?

Select up to 5 models from the list above to see their efficiency radar chart comparison.

Use the quick comparison presets or search for specific models to get started.