Skip to content
NewsModel release

Grok 4.5 canary string surfaces in xAI web UI, signaling imminent launch of 1.5T-parameter V9 model

· by Pondero Newsdesk

The short version

A UI upgrade string reading 'Unlock the full power of Chat with Grok 4.5' appeared in the Grok web client on July 6, 2026. xAI has not made a formal announcement, but the signal matches the company's pattern of UI-string-to-launch gaps of one to three days.

Grok 4.5 canary string surfaces in xAI web UI, signaling imminent launch of 1.5T-parameter V9 model

The text "Unlock the full power of Chat with Grok 4.5" appeared inside the Grok web client's account upgrade page on July 6, 2026, spotted by TestingCatalog and independently documented by Kie.ai. xAI has not issued a formal announcement. The same pattern preceded Grok's prior model launches, with UI strings typically appearing one to three days before the official post.

What

The canary string was embedded in the account upgrade flow, not in a system prompt or release note. A matching prompt appeared in the Grok Build CLI, directing users to the Heavy plan upgrade page, per Kie.ai's writeup published July 7.

The model behind the string is built on xAI's V9 foundation with 1.5 trillion parameters, roughly three times the scale of the existing v8-small model at approximately 0.5 trillion parameters, according to a breakdown by researcher Rohan Paul cited by Kie.ai. The Cursor team contributed supplemental fine-tuning and reinforcement learning data on top of the V9 base, adding coding-focused capability without altering the pre-training run.

Separately, Elon Musk posted on June 28 that Grok 4.5 had entered private beta at SpaceX and Tesla. He described its performance as "close to or beyond Opus" in those internal evaluations, per TestingCatalog's reporting. No benchmark scores, context window figures, or API pricing details have been published. Musk's performance claim is self-attributed and has not been verified against a third-party evaluation.

On July 6, the same day the canary string appeared, version numbers were reportedly scrubbed from the Grok interface menus, a step that observers noted preceded the last xAI model launch.

Why it matters

For developers and teams currently choosing between frontier coding models, Grok 4.5's Cursor SFT and RL training is the detail worth watching. Cursor's involvement in fine-tuning sits alongside xAI's existing Grok 3 integration in Cursor itself, and a stronger V9-based model could shift cost-performance comparisons against Claude Sonnet 5 and GPT-5.6 Terra on agentic coding tasks.

The 1.5T parameter count makes this the largest model xAI has shipped publicly. Scale alone does not guarantee capability gains, and xAI has not released any third-party evaluation results. Until benchmarks are published, the "close to or beyond Opus" characterization should be treated as a vendor-side impression, not a verified ranking.

The Cursor fine-tuning layer adds another dimension. The training data mix matters as much as the base model size for coding tasks, and xAI has not disclosed the scope or composition of the Cursor dataset that was added. Developers will need to wait for hands-on access to form an independent view.

Context

xAI has run a versioned foundation-model track since at least late 2025. The V9 run at 1.5T is not the end of that progression. A separate 2T-parameter V-series run was reportedly underway as of late June, with a finish date in late July and an expected August release, per Kie.ai. That would make Grok 4.5 a mid-cycle release rather than the top of the current architecture family.

xAI's official news page had no Grok 4.5 announcement as of the morning of July 8, 2026. Access to the page returned a 403 response during verification for this article, consistent with the page requiring authentication; the most recent publicly confirmed post predates the July 6 canary string.

What to watch next

The formal xAI announcement is the first gate. Once the model ships publicly, the key unknowns are API pricing relative to GPT-5.6 Terra and Claude Sonnet 5, any published benchmark comparisons on coding and agentic tasks, and whether Grok 4.5 lands in the Grok Heavy subscription tier immediately or phases in. The separate 2T run targeting August will set the ceiling for xAI's current architecture cycle.

Sources