Musk's SpaceXAI Rolls Out Grok 4.5, Pitches It as a Cheaper Rival to Anthropic's Opus

SpaceXAI put out Grok 4.5 on Wednesday, its first big model release since the company went public a few weeks ago, and Elon Musk wasted no time framing it as a direct challenger to Anthropic's top-tier model.

In a blog post announcing the release, SpaceXAI positioned Grok 4.5 as a general-purpose workhorse built to handle the tasks the AI industry has been racing to automate for years: coding, app development, office and clerical work, research, and writing. Nothing exotic here — the pitch is competence across the board rather than a narrow specialty.

The more interesting claim is on cost. SpaceXAI says Grok 4.5 delivers "twice greater token efficiency" than other leading models. Tokens are the chunks of text AI models process and generate, and how many of them a task burns through is directly tied to what it costs to run. Efficiency claims like this matter a lot right now, because token costs have become one of the biggest pain points for companies building on top of these models. If SpaceXAI's number holds up outside of marketing copy, it's a genuine selling point.

The company backed up the pitch with benchmark scores comparing Grok 4.5 to rival models. According to SpaceXAI's own figures, Grok is competitive with the best models on the market today, though it falls just short of best-in-class on some measures.

Musk's Opus comparison

Musk took to X, the social platform SpaceXAI owns, to make the comparison explicit, invoking Opus, Anthropic's model built for heavy, complex reasoning work.

"Based on strong positive feedback from customers in our beta test program, @SpaceXAI will make Grok 4.5 available to the public tomorrow. It is an Opus-class model, but faster, more token-efficient and lower cost," Musk wrote.

He followed up with a more specific claim: "Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster. The combination of capability, faster speed and lower cost is what makes it competitive."

That's a notable thing to say out loud, given Opus has been treated as something of a gold standard for demanding, high-complexity tasks. Musk isn't claiming Grok beats it outright — he's claiming Grok gets close while costing less and running faster, which is a different kind of pitch, and arguably a more practical one for businesses watching their AI bills.

The price comparison

The numbers back up at least part of that argument. SpaceXAI priced Grok 4.5 at $2 per million input tokens and $6 per million output tokens. Opus 4.7 costs $5 per million input tokens and $25 per million output tokens by comparison — meaning Grok undercuts it substantially, especially on output, which is often where costs stack up fastest on long or complex responses.

OpenAI's pricing sits across a range depending on which model tier customers use. Its most expensive model, Sol, runs $5 per million input tokens and $30 per million output tokens. Its cheapest, Luna, runs $1 per million input tokens and $6 per million output tokens. So Grok 4.5 lands in an odd but strategically useful spot: pricier than OpenAI's budget option, but promising more capability for roughly the same output cost, and dramatically cheaper than either company's flagship-tier offering.

A crowded week for model launches

The timing isn't accidental, or at least it's convenient. OpenAI is planning to release GPT 5.6 on Thursday, calling it its "strongest model yet." That release had reportedly been held up previously by the Trump administration over security concerns tied to the model's capabilities, according to the reporting. Now it's arriving just a day after Grok 4.5, putting two of the industry's most closely watched labs on stage in the same 48-hour window.

For SpaceXAI, the stakes go beyond just the usual model one-upmanship. This is the company's first major release since going public, which means Grok 4.5 also functions as a signal to investors and enterprise customers about where the company stands technically now that it's operating under public-market scrutiny. Pitching a model as "Opus-class" while undercutting Opus on price is a clean, easy-to-repeat message — the kind that's built for headlines and sales calls as much as for technical benchmarks.

Whether Grok 4.5 actually performs like an Opus-class model in the kind of messy, real-world use cases enterprises throw at these systems is something only time and independent testing will settle. Benchmarks published by the company that built the model are a start, not a verdict. But the pricing gap alone gives SpaceXAI a real argument to make to any customer currently watching their token bill climb — and with GPT 5.6 landing right behind it, that argument is about to get tested against fresh competition almost immediately.