Grok Build 0.1: Prix de niveau supérieur, Intelligence de niveau intermédiaire, Verbosité brutale

Grok Build 0.1 0616 costs $1.00 per million input tokens and $2.00 per million output tokens. That makes xAI's newest proprietary model the fourth cheapest out of 155 models on the Artificial Analysis platform, undercutting the average input price of $1.50 and demolishing the average output price of $8.00.

Intelligence Score Sits Above Average, Below the Elite

xAI's model scores a 40 on the Artificial Analysis Intelligence Index (v4.1, which rolls up GDPval-AA v2, Terminal-Bench v2.1, SciCode, GPQA Diamond, and others). That's 11 points above the class average of 29, landing it at #27/155. Not a top-10 contender, but solidly in the upper quartile for raw reasoning and knowledge work.

Speed Is Decent, But Verbosity Is a Hidden Tax

The model pushes 93.3 output tokens per second, a hair above the 86 average. Fine for interactive use. The real gotcha: verbosity. To complete the Intelligence Index evaluation, Grok Build 0.1 generated 130 million output tokens. That's 40% more than the average of 93 million. At $2.00/M output tokens, each full evaluation costs $375.01. You pay for that talkativeness.

Context, Reasoning, and Practical Takeaway

It handles 256k token contexts, accepts text and images, and outputs text. xAI labels it a reasoning model, and the benchmarks confirm it can chain logic. For cost-sensitive tasks like batch classification or mid-complexity agent workflows, Grok Build 0.1 offers a compelling price-to-performance ratio - as long as you budget for its windy responses. Expect xAI to iterate on verbosity in future builds, because at current output rates, the token bill adds up fast.

Source: Grok Build 0.1: Intelligence, Performance and Price Analysis
Domain: artificialanalysis.ai