SpaceXAI releases Grok 4.5, which is now an Opus-class model

Ronni Holmvig Strøm · 2026-07-09

SpaceXAI released Grok 4.5 on Wednesday, its first model since the company went public, and Musk reached for the comparison everyone expected. "It is an Opus-class model, but faster, more token-efficient and lower cost," he [wrote on

The Price Is the Argument

Grok 4.5 costs $2 per million input tokens and $6 per million output. Anthropic's Opus 4.7, the model Musk named, costs $5 and $25. The same class of capability at a fraction of the bill.

Look at where the price actually lands, though. $2 in, $6 out is almost exactly OpenAI's cheapest tier, Luna, at $1 and $6. xAI has priced a model it calls frontier-grade at the level of a budget one. The claim goes further than undercutting Opus. Opus-class capability now costs what commodity capability costs.

SpaceXAI's own benchmarks tell a more modest story. The charts published Wednesday show Grok competitive with the top models from OpenAI and Anthropic, and, in the company's own framing, just short of best-in-class. Capability Converges, Competition Moves

When several labs can all reach roughly the same frontier, the frontier stops being where the contest happens. 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."

The efficiency claim carries more weight than any single benchmark. SpaceXAI says Grok 4.5 has "twice greater token efficiency" than other leading models, and token efficiency is what turns a capable model into a usable one once you are paying by the million. A model that answers in half the tokens is cheaper twice over: once on the sticker price, again on every call. If that number holds outside xAI's own harness, it matters more than a two-point benchmark gap.