Grok 4.6 sits on the cost-versus-intelligence line, and that is the figure that matters for an office deciding what to pay.

On August 12, Artificial Analysis published independent scores placing Grok 4.6 at 61 on the Intelligence Index, tying GPT-5.6 Sol at max effort. Claude Opus 5 scored 63 and Claude Fable 5 scored 62. Grok 4.5 had scored 56.

The more useful number for a monthly bill is the other axis. Artificial Analysis measured about $0.84 per task and placed Grok 4.6 on its intelligence-versus-cost-per-task Pareto frontier. List pricing remained $2 per million input tokens and $6 per million output tokens, compared with $5 and $30 for GPT-5.6 Sol and $5 and $25 for Claude Opus 5.

Long agent work follows the same pattern. On AA-Briefcase, Grok 4.6 posted an Elo of 1577 and finished those tasks in about 53 turns and about 0.5 billion input tokens on average. Claude Opus 5 at max effort took about 103 turns and about 2.0 billion input tokens.

Brownsboro businesses that buy AI by the month should compare finished-task cost rather than brand name alone. A cheaper list price that takes twice as many turns can still cost more, so an independent cost-per-task number is a better starting point than a marketing slide. One caveat belongs in the fine print: rates can rise when a single request carries a very large prompt.

My I.T. Consultants is a managed service provider in Athens, TX. For a short look at which AI tools your office is already paying for, call 903-675-5431, Monday through Friday, 7:00 a.m. to 6:00 p.m.