xAI plans back-to-back Grok 4.6 and 4.7 releases at record cadence

xAI's strategy is velocity: releasing Grok 4.6 and then 4.7 in rapid succession (reportedly roughly two and four weeks out) following its July Grok 4.5 launch — a cadence xAI touts as unmatched by any frontier lab. Grok 4.6 is described as a 2-trillion-parameter model that has completed initial pre-training and could exceed Moonshot's Kimi K3 while running faster and more token-efficiently, positioning it directly against the Chinese open-weight challenger rattling U.S. labs.

Beyond the models, xAI is pushing into distribution and productivity surfaces. It launched a Microsoft Excel Grok plug-in — an unusual enterprise-integration move for the company — and deployable applications for Grok Build. Elon Musk promoted Grok Build directly, telling users to 'download Grok Build and type /tutorial,' signaling a consumer-app push alongside the model releases.

The rapid-fire versioning is a double-edged sword. It keeps xAI in the news and feeds the perception of momentum, but it also risks 'model-version-fatigue' — a theme running through the week as vendors ship incremental point releases faster than anyone can evaluate them. The competitive subtext is Kimi K3: xAI, like the broader U.S. industry, is racing to demonstrate that Western models can out-run a cheap, capable Chinese open-weight model on both quality and efficiency. Whether the 4.6/4.7 claims hold up under independent benchmarking — and whether an Excel plug-in meaningfully expands Grok's reach — are the open questions.