State of the Art
Top AI models across every dimension. Data from Artificial Analysis.
Last updated: May 3, 2026
Smartest
Highest intelligence scores across all benchmarks
1
GPT-5.5 (xhigh)OpenAI
602
GPT-5.5 (high)OpenAI
593
Claude Opus 4.7 (max)Anthropic
574
Gemini 3.1 Pro PreviewGoogle
575
GPT-5.5 (medium)OpenAI
57Fastest
Highest output throughput in tokens per second
1
Mercury 2Inception
782 tok/s2
Granite 4.0 H SmallIBM
461 tok/s3
Qwen3.5 0.8BAlibaba
343 tok/s4
Gemini 3.1 Flash-Lite PreviewGoogle
341 tok/s5
Qwen3.5 2BAlibaba
330 tok/sBest Value
Highest intelligence-to-price ratio
1
Qwen3.5 0.8BAlibaba
500.02
Qwen3.5 4BAlibaba
450.03
Qwen3.5 4BAlibaba
383.34
Qwen3.5 2BAlibaba
375.05
Qwen3.5 9BAlibaba
290.9Cheapest
Lowest cost per million tokens
1
Qwen3.5 0.8BAlibaba
$0.02/M tok2
Qwen3.5 2BAlibaba
$0.04/M tok3
LFM2 24B A2BLiquid AI
$0.05/M tok4
Qwen3.5 4BAlibaba
$0.06/M tok5
Qwen3.5 4BAlibaba
$0.06/M tokLowest Latency
Fastest time to first token
1
NVIDIA Nemotron 3 NanoNVIDIA
0 ms2
Ministral 3 3BMistral
0 ms3
Grok 3 mini Reasoning (high)xAI
1 ms4
LFM2 24B A2BLiquid AI
1 ms5
Qwen3.5 9BAlibaba
1 ms