SpaceXAI launched Grok 4.5 today, July 9, 2026, positioning it as an Opus-class model that trades capability for speed and lower cost. Elon Musk said it on X first: “It is an Opus-class model, but faster, more token-efficient and lower cost.”
Grok 4.5 is SpaceXAI’s first release since the company acquired Cursor, the AI coding tool. The model was trained with Cursor. It shows—the new release is pitched as a coding and agentic-work tool, not a consumer chatbot.
Availability and Pricing
Grok 4.5 is live in Grok Build, in Cursor on all plans, and via the SpaceXAI console. It’s not available in the EU yet. Pricing is $2 per million input tokens and $6 per million output tokens.
That puts it slightly cheaper than OpenAI’s GPT-5.6 Terra ($2.50 in, $15 out) on input but more expensive on output. The split matters for different use cases. Output-heavy work like report generation or code generation favors OpenAI’s pricing. Input-heavy work like search or retrieval favors Grok.
Training and Performance
SpaceXAI says Grok 4.5, trained alongside Cursor, outperforms comparable models on engineering and knowledge work. The company is pitching it as “maximally truth-seeking,” a dig at competitors they see as overly cautious.
This is the standard positioning for new models in the AI arms race. Each lab claims their model is smarter, faster, cheaper, or more ethical than the others. Benchmark scores often confirm one or two of those claims. Real-world performance is noisier.
Context and Timing
Grok 4.5 ships the same day OpenAI rolls out GPT-5.6 to everyone. For the first time in months, the three major frontier labs—OpenAI, Anthropic, and xAI—all have publicly available models at roughly the same time. This is what competition looks like at the frontier.
SpaceXAI is marketing Grok 4.5 as the tool for real work, not chat. We’ll learn from real usage whether that positioning holds.
