Elon Musk posted on X this week that xAI’s next two AI models are on a rapid timeline, with Grok 4.6 targeting a release around August 7 and a larger Grok 4.7 following a few weeks after that. The announcement points to an accelerating release pace among frontier AI labs, with xAI now positioned to ship three major model versions in roughly five weeks.
What Musk Revealed
Musk said Grok 4.6 will be a 1.5-trillion-parameter model featuring significantly improved supervised fine-tuning and reinforcement learning. Grok 4.7 will scale up further to 2.1 trillion parameters. “This will be better than 4.6 in every way, except slightly slower to serve, albeit with even better token efficiency,” Musk wrote on X.
Both dates are targets rather than firm commitments, and xAI hasn’t disclosed pricing or availability details for either model. Musk’s own track record on stated AI timelines has slipped before, and neither claim came with a benchmark sheet or technical paper, so the specifics should be treated as company statements rather than independently verified facts for now.
xAI’s Rapid Cadence and the Competitive Field
The announcement follows the public launch of Grok 4.5 on July 8, a 1.5-trillion-parameter model built on xAI’s V9 foundation, priced at $2 per million input tokens and $6 per million output tokens. Musk had described Grok 4.5 as an “Opus-class” model, and more recently claimed in a July 25 post that Grok 4.5 sits alongside Anthropic’s newly released Opus 5 alone on what he called the “Pareto frontier” of AI models, a claim made without an accompanying chart or methodology from either company.
If Musk’s timeline holds, xAI will have shipped three major model versions in roughly five weeks, a pace that outstrips most rival labs. The releases arrive amid intensifying competition: Moonshot’s Kimi K3, a roughly 2.8-trillion-parameter model, is already shipping, Anthropic recently released Opus 5, and OpenAI’s GPT-5.6 Luna is priced at $1 per million input tokens.
Musk has generally framed xAI’s strategy around speed and cost efficiency rather than sheer scale, though the jump to 2.1 trillion parameters for Grok 4.7 suggests the company is now pursuing both at once.
The rapid pace comes as US officials keep closer tabs on frontier model releases generally. OpenAI and Anthropic have already agreed to give US officials early access to unreleased frontier models for safety testing, an arrangement that now extends to other major developers, including xAI, Google DeepMind, and Microsoft, under a voluntary federal review framework formalized by a June executive order.
Participating companies can share covered models with the federal government for up to 30 days before wider release, though it isn’t yet clear how that timeline interacts with Musk’s stated August targets for Grok 4.6 and 4.7.
Quick Links:
Comments
Be the first to leave a comment.