Elon Musk announced on June 28, 2026, that Grok 4.5 has entered private beta testing at SpaceX and Tesla. The model is built on xAI's 1.5-trillion-parameter V9 foundation model, which completed training on May 26, 2026, and incorporates Cursor IDE session data in supplemental training to sharpen coding and technical reasoning performance.
According to Musk, early internal evaluations show performance "close to, perhaps exceeding Opus" — referring to Anthropic's Claude Opus 4.8, which currently leads the Artificial Analysis Intelligence Index at 61.4. No independent third-party benchmarks have been released, making the claim unverified by external evaluators.
xAI has committed to a monthly release cadence of V9-based model variants through the end of 2026. Each iteration will be trained from scratch, a development cycle faster than any other frontier lab has publicly pledged. SpaceX and Tesla serve as production evaluation environments: SpaceX's aerospace engineering workflows and Tesla's vehicle software development provide real-world technical tasks that Musk argues are harder than any benchmark.
Looking further ahead, Grok 5 is targeting 6 to 10 trillion parameters — potentially the largest model architecture ever publicly discussed — and is currently training on the Colossus 2 supercomputer alongside six other concurrent runs.
The competitive implications are significant. xAI is iterating faster than other frontier labs, using real production environments at Musk-owned companies as evaluation infrastructure. Whether monthly training cycles at this scale produce compounding capability gains is the open question for the second half of 2026.
