xAI releases Grok 4.6, matching frontier intelligence at 85% lower cost than Claude

xAI releases Grok 4.6, undercutting Claude on price and matching frontier benchmarks

xAI released Grok 4.6 claiming to match Claude Opus-class intelligence while pricing at an 85% discount—80% cheaper on input tokens and 88% cheaper for output tokens. The model builds on additional mid- and pre-training from the Grok 4.5 checkpoint, paired with newer supervised fine-tuning stages using Grok 4.5 traces and model-based filtering.

The release marks the sharpest pricing move yet in the model wars. Meta began the trend, and xAI is now pressing it harder, effectively forcing competitors to choose between margin and market share.

Grok Bot

Alongside the model release, xAI launched Grok Bot, an AI agent product that signs into tools and completes work autonomously—returning finished output rather than drafts or suggestions. This positions Grok as an agentic work product, not just a chat interface.

Benchmark and adoption gap

Grok 4.6 scores 61 on the AI Index, placing it at frontier level. Yet adoption remains slower than Cursor's, despite the performance parity. The hosts note that benchmark trust is at an all-time low—enough releases have shown strong numbers followed by weaker real-world experience that the gap between test scores and lived experience now weighs heavily on credibility.

Market momentum

SpaceX stock, which owns a stake in xAI, rose nearly 40% over five days following the announcement, including a 10% single-day pop. The hosts attribute some of the rise to semiconductor sector thesis momentum, though causality is unclear.