xAI Opens Grok 4.5 To The Public With Cursor-Trained Coding Push

xAI made its Opus-class Grok 4.5 model publicly available on July 9, trained on tens of thousands of NVIDIA GB300 GPUs and jointly tuned with Cursor for coding and agentic workflows.

Key Takeaways

  • xAI made Grok 4.5 publicly available on July 9, 2026, described by Elon Musk as an Opus-class model that is faster, more token-efficient and lower cost than comparable frontier systems.
  • Grok 4.5 runs on xAI's 1.5-trillion-parameter V9 architecture, trained on tens of thousands of NVIDIA GB300 GPUs with supplemental data from Cursor to boost coding, science, engineering and math performance.
  • The model launched the same day as OpenAI's GPT-5.6 Sol and Terra, the first day three frontier models went generally available on the same date.
  • Before public release, Grok 4.5 ran a closed beta at SpaceX and Tesla for embedded firmware and manufacturing code review, and Cursor offers native support in its editor from day one.
  • Developers can access Grok 4.5 via the xAI Console, Grok Build and the AI SDK, joining GPT-5.6, Anthropic's Fable 5 and Z.ai's GLM-5.2 as four Opus-class options within eight weeks.

xAI Opens Grok 4.5 To The Public With Cursor-Trained Coding Push

xAI's Grok 4.5 became publicly available on July 9, 2026, ending a short private beta with SpaceX, Tesla and Cursor. Elon Musk described the release as an Opus-class model that is faster, more token-efficient and lower cost than comparable frontier systems.

1.5-trillion-parameter V9 foundation, GB300-trained

Grok 4.5 runs on xAI's V9 architecture — a 1.5-trillion-parameter foundation trained across tens of thousands of NVIDIA GB300 GPUs. The training run pulled in supplemental data supplied by Cursor, which xAI says biased the model toward high-quality coding, science, engineering and mathematics performance. Developers can access it through the xAI Console, Grok Build, and the AI SDK.

Coding and agentic head-to-head with GPT-5.6

The public launch lands within hours of OpenAI's GPT-5.6 Sol rollout, marking the first day in AI history that three frontier models — GPT-5.6 Sol, Terra, and Grok 4.5 — went generally available on the same date. xAI is positioning Grok 4.5 as its most capable model for coding and agentic tool use, with strong Terminal-Bench and SWE-Bench scores, native availability inside the Cursor editor, and Cursor-optimised prompt formats.

Frontier AI model race

From SpaceX and Tesla beta to public API

Ahead of the public launch, Grok 4.5 spent several days in a closed beta at SpaceX and Tesla, where it was used for embedded firmware and manufacturing code review. Musk teased on X on July 8 that beta feedback was strong enough to accelerate the timeline. Cursor CEO Michael Truell confirmed native support in the editor from day one, giving Grok 4.5 an immediate distribution channel among professional developers.

The model arrives as U.S. hyperscalers publish record capex for accelerators and as the White House's cyber Executive Order compresses time-to-market for frontier releases. Between GPT-5.6, Grok 4.5, Anthropic's Fable 5 and Z.ai's GLM-5.2, buyers now have four Opus-class options within an eight-week window.

Reporting based on coverage from xAI, Bloomberg, Bloomingbit and Cursor.

Category: Machine Learning

Tags: Machine Learning AI Models AI OpenAI Nvidia

Related Articles

Frequently Asked Questions

When did Grok 4.5 become publicly available?

Grok 4.5 launched publicly on July 9, 2026, following a short private beta with SpaceX, Tesla and Cursor. Musk said strong beta feedback accelerated the timeline.

What hardware and architecture power Grok 4.5?

It is built on xAI's V9 architecture, a 1.5-trillion-parameter foundation trained across tens of thousands of NVIDIA GB300 GPUs, with Cursor-supplied data biasing it toward coding, science, engineering and mathematics.

How can developers access Grok 4.5?

Grok 4.5 is available through the xAI Console, Grok Build and the AI SDK, and it is natively supported inside the Cursor editor from day one with Cursor-optimised prompt formats.

How does Grok 4.5 compare to other frontier models?

xAI positions it as its most capable model for coding and agentic tool use, citing strong Terminal-Bench and SWE-Bench scores. It launched the same day as GPT-5.6 Sol and Terra, and joins Anthropic's Fable 5 and Z.ai's GLM-5.2 among four Opus-class options released within eight weeks.