xAI's Grok 4.6 Reaches Intelligence Frontier, Matching GPT-5.6 Sol on Third-Party Benchmarks

My take: Grok 4.6's official release on August 12 closes the question we had left open: yes, the model reaches the intelligence frontier according to independent evaluators. Artificial Analysis gave it a score of 61 on its Intelligence Index, the same level as OpenAI's GPT-5.6 Sol and five points above Grok 4.5. Claude Opus 5 (63) and Claude Fable 5 (62) remain ahead, but the gap is narrow.

The most relevant factor for anyone working with frontier models is the price: $2 per million input tokens and $6 per million output tokens makes Grok 4.6 one of the most cost-competitive options at that capability level. If you already use xAI in your workflow, this update has a direct use case.

A note of caution: xAI published an ELO of 1,753 citing the Databricks Leaderboard, but the independent LMArena places it at 1,464 with only 2,448 votes flagged as preliminary. Both platforms are third parties, but the data is still accumulating. The Artificial Analysis Index score (61) is the most consolidated and reliable number available today.

Have you evaluated whether a frontier model at lower cost could improve your team's efficiency?