SpaceXAI releases Grok 4.5

SpaceXAI has released Grok 4.5, a new AI model that Elon Musk describes on X as “Opus-class” while being faster and cheaper than rival frontier systems. According to TechCrunch, it is the company’s first model launch since SpaceXAI went public several weeks ago, and the company says it will be broadly available the day after Musk’s post. The Decoder reports the model is already accessible via Grok Build, Cursor, and the xAI console, with EU availability targeted for mid-July.

The immediate disruption is not a benchmark leap but a pricing move. SpaceXAI is charging $2 per million input tokens and $6 per million output tokens, TechCrunch reports, versus $5 and $25 respectively for Anthropic’s Opus line. The Decoder frames it as a “close-enough” strategy: if Grok 4.5 lands near the top tier on common coding and agentic benchmarks, the marginal performance gap may matter less than the cost difference for developers running large workloads. SpaceXAI also claims “twice the token efficiency” compared with other leading models, and The Decoder cites xAI saying Grok 4.5 uses fewer tokens than Opus on SWE Bench Pro tasks—an assertion that, if borne out in real usage, compounds the headline price cut.

That combination—lower list price plus claimed token efficiency—targets the buyers who actually feel the bill: startups building agent workflows, enterprises running internal copilots, and tool vendors embedding models into products with thin margins. It also pressures competitors that have grown used to selling “frontier” as a premium category, where the price is partly justified by scarcity—scarce compute, scarce talent, scarce safety review capacity. If a newly public company is willing to undercut that tier while advertising comparable capability, it forces a choice on incumbents: match pricing and accept lower margins, or keep pricing and risk being treated as a luxury option.

The release lands amid a crowded calendar. TechCrunch notes OpenAI plans to release GPT 5.6 on Thursday and describes it as its “strongest model yet,” after what the outlet says had been earlier limits tied to Trump administration security concerns. In other words, the market is being asked to compare two different kinds of advantage in the same week: one vendor selling “strongest,” another selling “fast enough, now cheaper.”

SpaceXAI’s blog post calls Grok 4.5 a workhorse for coding, app-building, office tasks, research, and writing. The first real test will be whether developers keep paying for “best-in-class,” or quietly move their token-heavy workloads to the model that makes the invoice smaller.