xAI's new Grok 4.5 claims to code using 4x fewer tokens than rivals

Elon Musk's xAI has launched Grok 4.5 — the strongest model in its lineup, built specifically for coding, agentic tasks, and knowledge work. The release landed on July 16, 2026.

A Focus on Coding and Agentic Tasks

Grok 4.5 was trained alongside developers at Cursor, the popular AI-powered coding environment, on datasets spanning coding, science, engineering, and math. According to xAI, the model handles challenging Rust and C/C++ tasks confidently and can build entire applications end-to-end, from prompt to production, in a single pass.

Speed and Efficiency

The model runs at speeds of about 80 tokens per second and, according to xAI, is twice as token-efficient as leading rival models on comparable tasks. In office and knowledge-work scenarios, Grok 4.5 resolves tasks using roughly 4.2x fewer output tokens than competing models — the model is capable of building complex Excel models that involve researching data from the web and formulas linking multiple sheets.

Pricing and Availability

Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens. It's available in the Grok Build environment, across all Cursor plans, and through the xAI console.

Why It Matters

Grok 4.5's positioning highlights a broader shift in competition among frontier AI models: the race is no longer just about absolute answer quality, but about efficiency — how many tokens, and therefore how much money, it takes to solve the same real-world engineering task. For developers and companies that have already deeply integrated AI into daily work — from writing code to building financial models — that metric is increasingly becoming as decisive a factor in model choice as raw output quality.

This material is for informational purposes only and is not investment advice.