SpaceXAI releases Grok 4.5, which Elon describes as an ‘Opus-class model’

SpaceXAI has launched its newest mannequin, Grok 4.5 — the primary because the firm went public a number of weeks in the past.

In a weblog publish revealed Wednesday, SpaceXAI characterised its new launch as a workhorse that may sort out all the typical duties that the AI trade has sought to automate: coding and app-building, workplace and clerical work, analysis, writing, and different types of routine data work.

Grok can supposedly do all this for much less spend, too, as SpaceXAI says that its mannequin has “twice larger token effectivity” than different main fashions. If it carries via to real-world use instances, that effectivity can be a giant benefit for SpaceXAI, since the price of tokens has been a rising concern for AI customers.

The corporate launched benchmark metrics Wednesday that appeared to indicate Grok’s competitiveness with different prime fashions from SpaceXAI opponents, though simply in need of best-in-class:

In a publish on his social media platform X (which is a subsidiary of SpaceXAI), founder Elon Musk in contrast the mannequin to Opus, Anthropic’s LLM designed for intensive and sophisticated duties.

“Primarily based on robust constructive suggestions from prospects in our beta take a look at program, @SpaceXAI will make Grok 4.5 accessible to the general public tomorrow. It’s an Opus-class mannequin, however sooner, extra token-efficient and decrease price,” wrote Musk in his publish on X.

Musk later added: “Our inside evaluation is that Grok 4.5 is roughly similar to Opus 4.7, however a lot sooner. The mix of functionality, sooner velocity and decrease price is what makes it aggressive.”

SpaceXAI says that its new mannequin prices $2 per million enter tokens and $6 per million output tokens. That’s fairly aggressive, if Grok’s capabilities match SpaceXAI’s rhetoric.

Opus 4.7, by comparability, prices $5 per million enter tokens and $25 per million output tokens. OpenAI has tiered prices for various mannequin variations: Sol, its most costly, prices $5 for 1 million enter tokens and $30 for 1 million output tokens, whereas its least costly, Luna, prices $1 for 1 million enter and $6 for 1 million output tokens.

It’s a giant week for AI mannequin releases. OpenAI is planning to launch GPT 5.6, its newest, strongest mannequin, on Thursday. The discharge of that mannequin had beforehand been restricted by the Trump administration, as a result of considerations about its safety implications. OpenAI has known as it its “strongest mannequin but.”