Grok 4.5 arrives with a bold claim, top-tier ability at a bargain price

Elon Musk's AI operation shipped its biggest release of the year on July 8. Grok 4.5, from the newly public SpaceXAI, is pitched as the company's smartest model yet — and, notably, its first major release since the company acquired the AI coding startup Cursor.

The pitch: flagship quality, mid-tier price

Musk describes Grok 4.5 as an "Opus-class" model — a reference to Anthropic's long-standing top-tier family — but faster, more token-efficient, and cheaper. The pricing backs up at least the last part: $2 per million input tokens and $6 per million output tokens, which undercuts every comparable frontier model on the market. For comparison, OpenAI's new flagship Sol costs $5 in and $30 out.

Whether the capability claim fully holds will take a few weeks of independent testing to settle, but the strategy is unmistakable: SpaceXAI is competing on price-per-unit-of-intelligence rather than raw leaderboard position.

Built with and for coding

Grok 4.5 was trained on data spanning coding, science, engineering, and math, and the company says it was developed alongside Cursor — the popular AI coding editor SpaceXAI now owns. That relationship shows in the distribution plan: the model is available immediately in Cursor on all plans, in the company's Grok Build product, and through the SpaceXAI developer console.

The emphasis throughout is on agentic tasks and real engineering work — the model reasoning efficiently through practical problems rather than burning thousands of tokens thinking out loud.

One catch for some readers

Grok 4.5 is not yet available in the EU, with availability there expected in mid-July. Canadian users are unaffected, but businesses with European operations or EU-based staff should note the gap.

A crowded week

The release lands in the middle of the busiest stretch of AI launches this year, with OpenAI's GPT-5.6 family opening to the public, Meta shipping a new model, and Google's Gemini 3.5 Pro due within days. Buyers have not had this much simultaneous choice at the top end of the market before.

Why it matters for small business

The headline here is price competition at the top tier. Frontier-level AI at $2/$6 per million tokens changes the math for any business considering custom AI work — a developer building you a quoting tool or customer-service assistant can now use a top-shelf model without the top-shelf bill. If your business writes software, or hires people who do, the Cursor integration means Grok 4.5 shows up in a tool many developers already use, at no extra cost on existing plans. The sensible move this month is not to switch anything immediately — it is to ask whoever builds or supplies your AI tools which model they use, and whether this new pricing changes what you pay.