Elon Musk’s xAI is preparing to unleash a pair of massive AI models in rapid succession this summer, with a 1.5 trillion parameter Grok 4.5 already in private beta and a 2 trillion parameter follow-up targeting an August launch.
Musk announced on June 28 that Grok 4.5, built on what xAI calls its V9 foundation model, has finished training and is currently being tested internally at SpaceX and Tesla. No public API access or external benchmarks have been made available yet.
The numbers behind the next Grok generation
Here’s the thing about 1.5 trillion parameters: it represents roughly a threefold increase over the existing Grok 4.x production series, which runs on approximately 500 billion parameters.
The 2 trillion parameter model is reportedly in the final stages of training and expected to complete by late July. A public launch is anticipated for August, though xAI hasn’t committed to specific dates for broader access to either model.
Early internal evaluations suggest Grok 4.5 may rival or exceed Anthropic’s Claude Opus in capability. But without independent benchmarks, it remains an internal assessment rather than a verified fact.
xAI’s approach of deploying first within Musk-affiliated companies before any public rollout is notable. SpaceX engineers and Tesla teams create a testing environment that synthetic benchmarks simply can’t replicate.
The Colossus advantage
Powering this rapid development cycle is xAI’s Colossus supercluster in Memphis, Tennessee, which now houses over 200,000 NVIDIA GPUs. The company has ambitions to scale that infrastructure to one million GPUs.
Musk has outlined plans for xAI to release entirely new models, trained from scratch, on a monthly cadence through December 2026.