Musk plots back-to-back Grok 4.6 and 4.7 releases within weeks

In a series of posts, Elon Musk laid out an aggressive release cadence: 'Grok 4.6 releases around August 7. This will be the 1.5T model with significantly improved SFT & RL. Grok 4.7 will be the 2.1T model released a few weeks later.' (Other reports pegged 4.6 nearer 2 trillion parameters; the numbers vary across xAI's own channels.) Musk claimed 4.6 could 'exceed Kimi' — Moonshot's newly open-weighted K3 — while being faster and more token-efficient, with 4.7 better 'in every way, except slightly slower to serve.'

The cadence itself is the story: shipping two frontier model versions back-to-back within roughly a month is a pace no other frontier lab has matched, and it doubles as a data flywheel. xAI paired the roadmap with a Microsoft Excel Grok plug-in and deployable applications for Grok Build, explicitly targeting finance and coding workloads through deep Cursor integration — a direct play for the same enterprise coding revenue Anthropic and OpenAI are chasing.

Developer reaction was skeptical bordering on mocking. TechRadar's writeup of the Excel plug-in carried the headline 'for some reason,' capturing a broader sentiment that xAI is shipping breadth faster than it's proving depth. The version-fatigue theme runs through the week: Grok 4.6/4.7, Opus 5, Gemini 3.6 Flash, and Qwen3.8 Max all landing or teased within days.

What to watch: whether the August 7 date holds, whether 4.6 actually beats Kimi K3 on independent benchmarks (Musk's 'could exceed' is a claim, not a result), and whether the Cursor coding flywheel converts into paid seats. The token-efficiency claims will be the first thing developers stress-test.