SpaceXAI Releases Grok 4.5 on Wednesday, Pricing It Below Anthropic and OpenAI Rivals

Since OpenAI cleared government review for its GPT-5.6 line and set a Thursday public release, the AI model race has gotten considerably more crowded. xAI stepped into that same week with a model of its own.

xAI launched Grok 4.5 on Wednesday, making it available to the public on a limited basis the same day, according to a company statement reported by Forbes. The company is calling it its "strongest model ever."

What Grok 4.5 Is Built For

The model is designed around agentic tasks: work that AI can execute autonomously without a human steering each step. xAI says it performs well in finance, legal analysis, and coding, and can build complex Excel models using live web research, as well as generate diagrams and PowerPoint content, according to the company statement cited by Forbes.

TechCrunch noted the broader positioning: xAI is presenting Grok 4.5 as a workhorse for the full range of tasks the AI industry has been chasing, from coding and app-building to research and routine office work.

The Price Argument

Grok 4.5 costs $2 per million input tokens and $6 per million output tokens, according to TechCrunch.

For comparison: Anthropic's Opus 4.7 runs $5 per million input tokens and $25 per million output tokens. OpenAI's Sol, the most expensive tier of its GPT-5.6 family, costs $5 per million input and $30 per million output. OpenAI's Luna, its budget tier, costs $1 per million input and $6 per million output tokens.

Grok 4.5 sits comfortably below Opus 4.7 and OpenAI's Sol on price. Luna is cheaper on input tokens but matches Grok on output. xAI also claims twice the token efficiency of other leading models, which, if accurate in real-world deployments, would make the cost gap even larger in practice.

Musk's Own Framing

Elon Musk posted on X Tuesday that Grok 4.5 is "an Opus-class model, but faster, more token-efficient and lower cost." In a follow-up Wednesday, he wrote that xAI's internal assessment puts Grok 4.5 as "roughly comparable to Opus 4.7, but much faster."

X and xAI are separate but both Musk-controlled entities, so Musk is using his own platform to promote his own model. When weighing the self-reported benchmarks, this context matters.

The Benchmark Caveat

xAI released benchmark metrics Wednesday. According to TechCrunch, those numbers showed Grok 4.5 as competitive with top-tier models but not best-in-class. That is the company's own data. Independent third-party evaluations have not yet been reported.

AI companies routinely publish benchmarks that favor their own products. Anthropic and OpenAI do it too. Until developers run Grok 4.5 through real workloads and publish results, the "Opus-class" label is Musk's claim, not an established fact. xAI has clear financial incentive to position aggressively—the company went public several weeks ago, and a strong model launch supports that narrative.

The pricing structure is objective and already published. If the capability is even close to what xAI claims, $2/$6 per million tokens against Anthropic's $5/$25 is a real competitive advantage, not a marginal one. Cost pressure on AI tokens has been a growing concern for enterprise buyers, as TechCrunch noted. A model that is genuinely cheaper to run at comparable quality changes procurement math for large deployments.

Crowded Week

Grok 4.5 is launching into one of the most competitive single weeks the AI industry has seen. OpenAI's GPT-5.6—including the Sol, Terra, and Luna tiers—cleared government security review and is set for public release Thursday. The Trump administration had previously held up that release over national security concerns before approving it.

xAI chose to drop Grok 4.5 the day before OpenAI's launch. Whether that timing was strategic or coincidental, the effect is the same: buyers now have multiple major model releases to evaluate simultaneously.

The unresolved question is whether independent developers find Grok 4.5's agentic and finance capabilities as strong in practice as xAI's benchmarks suggest. If they do, the pricing alone makes it a serious challenger. If the real-world performance falls short of Opus 4.7, the cost advantage will not fully compensate.