Directories / Models / Value frontier

The model value frontier

Every scored model plotted by intelligence against what it costs. The line is the frontier: the smartest model at each price. Anything below it is paying more for less.

Full ranking board → What changed today

409 benchmarked models · Intelligence Index from Artificial Analysis · pricing OpenRouter-fresh

Intelligence per dollar

Up and to the left is better: smarter and cheaper. The frontier models are the ones worth defaulting to. Everything below them costs more for the same or less intelligence.

value frontier QwenOpenAIMistralGoogleMistralDeepSeekAnthropicMoonshotai

Best-value picks

6 on the frontier

Model	Intelligence	Blended $/M	In / Out $/M
Claude Fable 5 Anthropic	59.9	$20	$10 / $50
GPT-5.6 Sol OpenAI	58.9	$11.25	$5 / $30
Kimi K3 Moonshotai	57.1	$6	$3 / $15
GPT-5.6 Terra OpenAI	55	$5.63	$2.5 / $15
Grok 4.5 xAI	53.8	$3	$2 / $6
Claude Sonnet 5 (Free) Anthropic	53.4	$0	$0 / $0

Fastest models

tokens/sec

Mercury 2 1072 t/s Step 3.7 Flash (Free) 402 t/s Step 3.7 Flash 402 t/s Step 3.7 Flash 402 t/s Gemini 3.1 Flash Lite 333 t/s Nemotron 3 Nano Omni 30B A3B Reasoning 323 t/s gpt-oss-120b (free) 313 t/s gpt-oss-120b 313 t/s

Median output speed from Artificial Analysis. Higher is faster to stream a response.

The context-window race

top 12

X-Ai/Grok 4.1 Fast Reasoning 20M Llama 4 Scout 10M Grok 4.20 2M Grok 4.1 Fast 2M Grok 4 Fast 2M Grok-4-Fast-Non-Reasoning 2M Gemini 1.5 Flash 2M Auto Router (Beta) 2M Auto Router 2M Pareto Code Router 2M X-Ai/Grok-4-Fast-Non-Reasoning 2M X-Ai/Grok-4-Fast-Reasoning 2M