xAI's Grok 4.5 enters private beta at Tesla and SpaceX with 1.5T parameters

Elon Musk announced that xAI's Grok 4.5 has entered private beta inside SpaceX and Tesla operations. The model runs on a new V9 foundation architecture with 1.5 trillion parameters, a roughly 50% increase over Grok 4.4 released in May 2026. xAI says the model draws supplemental training from real-time data integrations, and reports suggest it incorporates data from AI coding platform Cursor.

Musk's internal evaluations reportedly indicate Grok 4.5 performs comparably to — or even surpasses — Anthropic's Claude Opus, though these are unverified first-party claims with no published third-party benchmarks. Deploying inside Tesla and SpaceX gives xAI a captive, high-signal testbed for agentic and real-time use cases before any public release.

The timing places Grok 4.5 directly against a crowded frontier: Anthropic's Sonnet 5 and Opus 4.8, OpenAI's GPT-5.6 family, and Google's Gemini line. xAI separately shipped state-of-the-art voice APIs via the Vercel AI Gateway, signaling a broader push into production developer tooling.

Skepticism is warranted. Musk's habit of touting internal wins without external validation means the "rivals Opus" claim should be treated cautiously until independent evals appear. The parameter-count leap is notable, but raw scale no longer maps cleanly to capability, and the private-beta framing means developers can't yet test the claims themselves. Watch for a public benchmark or an outside deployment to substantiate the hype.