Musk calls the new multimodal model comparable to Anthropic's top Claude family but more token-efficient. Private beta hit June 28; a wider public launch is set for July 9, at $2/$6 per million tokens.
2026-07-08 · by Moe Ameen
xAI — the Musk-founded AI company now branded SpaceXAI — released Grok 4.5, its new flagship model. It is a general-purpose reasoning model built for coding, agentic tasks, and knowledge work, with multimodal understanding (it reads text and images) and conversational use. The model entered private beta with SpaceX and Tesla teams on June 28, 2026, and xAI set a wider public launch for July 9; it is already reachable through the xAI/SpaceXAI console and API, inside Grok Build, and in Cursor on all plans. At launch it was not yet available in the EU.
Elon Musk described Grok 4.5 as an "Opus-class" model — roughly comparable to what was recently Anthropic's top Claude family — but faster, more token-efficient, and lower cost. A chart xAI published with the announcement claims the model outperforms Opus 4.8 on several benchmarks. Those are the company's own figures: xAI has not released a full technical report, and there are no independent evaluations yet, so the head-to-head should be read as a claim rather than a settled result.
API pricing is $2.00 per million input tokens and $6.00 per million output tokens, with configurable reasoning effort — low, medium, or high, defaulting to high — so callers can trade depth for speed and cost. Reporting ties the model to xAI's new V9 foundation, said to be around 1.5 trillion parameters and supplemented with training data from the Cursor coding platform, though the weights are closed and the parameter count comes from Musk's own statements rather than an independently verified technical report. The model accepts text and image input and returns text.
There are two ways a creator acts on this launch today, and being fast on both is the point of a news moment. The first is editorial: the "Opus-class, but cheaper" framing has made Grok 4.5 the model story of the week, and the frontier race is exactly what your audience is following right now. Drop your take — how Grok 4.5's claims stack up against Opus 4.8 and GPT, what $2/$6 pricing and tunable reasoning effort mean in practice, why "vendor benchmarks" deserve a raised eyebrow — into Kompozy as a source, and it fans that single point of view into a blog explainer, a comparison carousel, a few captioned short clips, and platform-native posts for X and LinkedIn, then schedules and publishes them across your channels. One clear opinion, published early, becomes a week of content.
The second is workflow, for the creators who will actually try the model: Grok 4.5 is a good place to reason and draft, but it stops at text. Use it to brainstorm angles and write rough scripts, then paste the best of it into Kompozy, which generates the media the model cannot — persona and avatar video, carousels, quote cards, infographics, blogs, and newsletters — in your brand voice via the Persona Brief, and publishes them to nine platforms plus email and blog on autopilot. Kompozy runs its own generation on managed Claude and OpenAI models, so there is nothing to wire; you draft in Grok, produce and ship in Kompozy. The model does the thinking; the engine does the shipping.
Grok 4.5 is xAI's new flagship model — a general-purpose reasoning model for coding, agentic tasks, knowledge work, and conversation, with multimodal understanding of text and images. It entered private beta on June 28, 2026, with a wider public launch set for July 9, and is reachable via the xAI/SpaceXAI API and console, Grok Build, and Cursor.
Elon Musk calls it "Opus-class" but faster, more token-efficient, and cheaper, and xAI published a chart claiming it beats Opus 4.8 on several benchmarks. Those are xAI's own numbers with no full technical report yet, so treat the comparison as a claim until independent evaluations land.
Via the xAI API it is $2.00 per million input tokens and $6.00 per million output, with configurable reasoning effort (low, medium, or high; high by default). Consumer access may also ride xAI subscription tiers.
No. It reasons over text and images and writes text; it generates no images, video, or posts and publishes nothing. To turn its drafts into finished, scheduled content, pair it with a content engine like Kompozy that produces the media and publishes across nine platforms plus email and blog.