Grok 4.5 Is Here: xAI's New Model Is Cheap, Fast, and Frugal With Tokens
xAI has released its new model, Grok 4.5 — built for coding, agentic tasks, and office work, and billed as the company's "strongest model yet." One interesting detail: it was trained alongside Cursor. But the part that stands out is less about performance than economics: xAI says Grok 4.5 solves the same work with far fewer tokens, and at a lower price, than rival models.
In brief
- What launched: Grok 4.5, for coding, agentic tasks, and knowledge work. Per xAI, its strongest model yet, trained alongside Cursor.
- Efficiency and price: $2 per million input tokens / $6 output; 80 tokens per second. xAI says that on one coding test (SWE Bench Pro) it finished the same work with about 4x fewer tokens than Opus 4.8.
- Benchmarks (xAI's numbers): near the top on coding tests but not first everywhere — leading on SWE Marathon, but a little behind Fable and GPT-5.5 on DeepSWE and Terminal Bench. Competitor scores are taken from those companies' own publications.
- Office + access: the default model in Grok Build; it can build complex files in Excel, PowerPoint, and Word. Available today in Grok Build, Cursor (all plans), and the API (
grok-4.5); not yet in the EU (expected mid-July). A limited-time free trial is available.
Our take
The real story for you isn't the leaderboard — it's the bill. Grok 4.5's pitch isn't to be the smartest model; it's to do the same work with far fewer tokens and for less money, which shows up directly in your monthly cost. And it's not just for coders: it handles ordinary office work — Excel models, slides, Word docs — which is most people's day job. Right now it's free for a limited time in Grok Build and Cursor, so if you have templated work you'd like to run cheaper, trying it costs nothing. Two caveats: the scores and token comparisons are xAI's own numbers, so don't take them on faith — and it isn't available in the EU yet (mid-July).
Kaynak: xAI