xAI just launched Grok 4.5, its first model since acquiring Cursor — built specifically for coding and agentic work, and priced well below Claude and GPT. Here's what actually changed.
Grok 4.5 Just Launched: What You Need to Know
Grok 4.5 landed on July 8, 2026, and it's not just another routine model update — it's xAI's first release since going public and acquiring the AI coding editor Cursor, and it's aimed squarely at developers rather than casual chatbot users.
What Grok 4.5 Actually Is
Unlike earlier Grok releases built as general-purpose chat assistants, Grok 4.5 was designed from the ground up for coding, agentic tasks, and knowledge work. xAI trained it using real developer session data pulled from Cursor, which the company acquired earlier this year — meaning the model learned from how people actually write and debug code in a live editor, not just from scraped code repositories.
Elon Musk described it as an "Opus-class" model, positioning it as a direct rival to Anthropic's higher-end coding models, but with a specific emphasis on speed and cost rather than raw benchmark supremacy.
How It Performs
On independent benchmarking, Grok 4.5 lands in a solid but not top-tier position — it ranks fourth on the Artificial Analysis Intelligence Index, ahead of every Gemini model and open-weight competitor, though behind Anthropic's top coding model and OpenAI's GPT-5.5 on several engineering benchmarks. Where it does stand out is efficiency: xAI reports it uses roughly four times fewer output tokens than Claude's Opus 4.8 to solve the same engineering tasks, and it leads on at least one real-world benchmark measuring sustained, multi-step coding work.
Speed and cost
Grok 4.5 currently runs at around 80 tokens per second, with xAI signaling that a custom inference stack in development could roughly double that speed without changing the underlying model.
Pricing: Where It Gets Interesting
This is the part most developers will actually care about.
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Grok 4.5 | $2 | $6 |
| Claude Opus 4.8 | $5 | $25 |
| GPT-5.5 (Sol, top tier) | $5 | $30 |
| GPT-5.5 (Luna, budget tier) | $1 | $6 |
At those rates, Grok 4.5 undercuts Claude's flagship coding model by more than 60% on output cost, while staying competitive with OpenAI's cheaper tier — a meaningful shift for teams running high-volume agentic workflows where token costs add up fast.
Where You Can Use It Today
Grok 4.5 is live now as the default model in Grok Build, available inside Cursor on all plans, and accessible through the xAI API console. One catch: it isn't available in the EU yet, with xAI targeting a mid-July rollout for that region.
What This Means for Developers
If your work leans heavily on agentic coding — long, multi-step tasks where an AI has to plan, write, test, and fix code across several rounds — Grok 4.5 is now a legitimate option to test alongside Claude and GPT-based tools, particularly if cost per task is a real constraint. It's not the outright top performer on every benchmark, but the combination of speed, price, and native Cursor integration makes it worth a trial run rather than a full switch on day one.
Frequently Asked Questions
Is Grok 4.5 better than Claude for coding?
It depends on the task. Independent benchmarks show Anthropic's top coding model still leads on most engineering tests, but Grok 4.5 is significantly cheaper and more token-efficient, and it wins on at least one real-world sustained-coding benchmark.
Can I use Grok 4.5 in Cursor right now?
Yes — it's available in Cursor on all plans as of launch, alongside xAI's own Grok Build tool.
Why isn't Grok 4.5 available in the EU yet?
xAI hasn't given a detailed explanation beyond regional rollout timing, but has stated it expects EU availability by mid-July 2026.
Is Grok 4.5 a general chatbot or a coding-specific tool?
It's built and trained primarily for coding, agentic tasks, and knowledge work, though it can still handle general questions like other Grok models.
Compare Grok, Cursor, and other coding assistants side by side in the AI tools directory.