What Is Grok 4.5? Coding Performance, Pricing, and the EU Caveat

What Is Grok 4.5 (New Model Overview)

First, let us establish what kind of model Grok 4.5 is. Its positioning and how it was trained both stand out.

Grok 4.5's Positioning (Coding-Focused)

Grok 4.5 is the latest AI model released by xAI. It claims high performance at coding, agentic tasks (where the AI breaks work into steps and does it automatically), and knowledge work such as research and document creation. xAI positions Grok 4.5 as its "strongest model ever."

Note that the official page labels the provider "SpaceXAI," while the footer copyright reads "xAI Corp." This article refers to the company as xAI. For the lineage from the previous generation, see also the Grok 4 explainer.

Trained Alongside Cursor

Another feature of Grok 4.5 is that it was trained alongside the AI code editor Cursor. Because it was honed together with a tool used in real development, the intent appears to be strength at practical, work-oriented tasks like coding and agentic work. Its training data is said to span knowledge in coding, science, engineering, and math.

Grok 4.5's Performance (Benchmarks)

Grok 4.5's performance is measured on software-development benchmarks. This section has a lot of numbers, so if you are not involved in development, you can take away just the conclusion—"its strengths are efficiency and price"—and read on.

Its Engineering Performance

Looking at the official benchmarks, Grok 4.5 sits on par with or ahead of leading models on software-development tests. On SWE Marathon, which measures long-running agentic work, it is first; on Terminal Bench 2.1 it is near the top; while on DeepSWE and others it places mid-pack—so it is not always the outright leader. That said, xAI claims it exceeds comparable leading models on real engineering tasks.

The model is said to have been trained across tens of thousands of NVIDIA GB300 GPUs, with training and stability techniques designed for large-scale runs.

Speed and Token Efficiency

Grok 4.5's biggest strength is speed and token (the unit in which AI processes text) efficiency. Its responses are fast at 80 tokens per second, and it uses fewer tokens to solve the same task. According to the official figures, Grok 4.5 resolves tasks with 15,954 tokens on average—about 4.2× fewer than Opus 4.8 (max), which uses 67,020. Fewer tokens means lower cost and less time.

Even at the same price, the fewer tokens you use, the lighter the actual bill. For how to think about comparing the cost of models, see also the Claude Opus cost comparison.

Grok 4.5's Pricing and Where You Can Use It

Next, let us confirm the pricing you are wondering about and where you can actually use it.

Pricing (Per Million Tokens)

Grok 4.5's API (the mechanism for calling it from an external program) pricing is $2 per million input tokens and $6 per million output tokens. On top of that, because it solves the same task with fewer steps and tokens, the real cost tends to be lower than the sticker price suggests.

Where You Can Use It (Grok Build, Cursor, API)

Grok 4.5 is the default model in xAI's app "Grok Build," and it is available on all plans in the AI code editor Cursor. Developers can also call it via xAI's API. Grok Build is said to handle spreadsheets, slides, and document creation too, extending beyond coding into office work. For the full range of Grok articles, including previous generations, browse the Grok article list.

Things to Watch When Using Grok 4.5

Finally, let us organize the caveats worth knowing before adoption. In particular, watch for regional differences in availability.

Not Yet Available in the EU

Easy to miss is the regional restriction. The official announcement clearly notes that Grok 4.5 is not yet available in the EU in any product or the API, with EU availability expected in mid-July.

What to Check Before Adopting

For regions outside the EU, such as Japan, there is no explicit statement from the official announcement. So before you actually use it, it is safest to check whether Grok 4.5 is selectable in your region and on your plan. All told, Grok 4.5 is a model that shines on price and efficiency, well suited to developers who want to do large volumes of coding and agentic work cost-efficiently. To compare it with other companies' latest models from the same period, see also the Claude Opus 4.8 explainer.

When working through official English-language documentation, converting a page to Markdown keeps the structure of headings and code intact, making the key points easier to follow. To reformat a web page as-is, the following tool is handy.

Free ToolURL to Markdown ConverterConvert any public web page URL to Markdown. Preserves headings, tables, lists, and links — perfect for LLM and RAG preprocessing, research notes, and archiving web articles.Try it now →