SpaceXAI releases Grok 4.5, which Elon describes as an 'Opus-class model'
TechCrunch ● Covered by 5 sources
SpaceXAI just dropped Grok 4.5, and Elon's calling it
Based on reporting by TechCrunch — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
SpaceXAI pushed out Grok 4.5 on Wednesday, its first big model release since the company went public, and the pitch is straightforward: this thing is built to grind through the same grunt work every AI lab is chasing right now — coding, spreadsheets, research memos, the usual pile of knowledge work that companies want automated. SpaceXAI says it's doing that with roughly double the token efficiency of rival models, which, if it holds up outside a benchmark slide, matters a lot given how much enterprises have started grumbling about token costs eating into their AI budgets.
Elon Musk didn't wait for reviewers to make the comparison himself, posting on X that Grok 4.5 is an "Opus-class model," referencing Anthropic's flagship line for heavy, complex reasoning tasks. He went further, telling followers the internal read is that Grok 4.5 sits roughly on par with Opus 4.7 — just noticeably faster. Musk's framing is capability plus speed plus price, not any single one of those in isolation, which is a fair way to sell a model that landed just short of best-in-class on SpaceXAI's own published benchmarks.
And the pricing does back up part of that story. Grok 4.5 runs $2 per million input tokens and $6 per million output tokens, undercutting Opus 4.7's $5 and $25. It also slots in below OpenAI's priciest tier, Sol, at $5 and $30, though OpenAI's cheaper Luna model, at $1 input and $6 output, still edges Grok on the input side. So SpaceXAI isn't claiming to be the absolute cheapest option on the market — it's arguing that at this capability level, nobody else comes close on cost.
The timing isn't an accident, either. OpenAI is set to launch GPT 5.6 on Thursday, a release the company is calling its strongest model yet, one that reportedly got held up earlier due to security concerns from the Trump administration. Two major model drops in two days, each company trying to frame itself as the value leader among frontier-capable systems — that's the state of the race right now, and SpaceXAI clearly wanted to get its narrative out first.
My take — AI-written commentary, not fact-checked reporting
Musk comparing his own model to a competitor's and having Musk himself do the benchmarking commentary is not exactly a neutral data point, so I'd wait for third-party evals before crowning Grok 4.5 anything. That said, the pricing gap versus Opus is real and meaningful, and it's a reminder that the actual AI price war is happening on inference cost, not headline IQ points. Everyone's converging on similar capability tiers, so cost-per-token is quietly becoming the metric that decides who actually gets deployed in production.
Read more about this at: TechCrunch
