Grok 4.5: X.ai's New AI Excels in Coding & Legal Tasks

Grok 4.5 is SpaceXAI's newest and most intelligent model, designed for coding, agentic tasks, and knowledge work. It was trained alongside Cursor and is described as their strongest model ever. What's interesting is Grok 4.5 excels at real engineering tasks and surpasses comparable leading models. It also scores number one on Harvey's Legal Agent Benchmark for office work. The model was trained on tens of thousands of NVIDIA GB300 GPUs, with a strong focus on data filtering and curation. Grok 4.5 is highly capable in coding, handling challenging Rust and C/C++ tasks, and building end-to-end applications from a single prompt. It can also create complex Excel models and meticulously build diagrams in PowerPoint and write clear prose in Word. The model operates at fast speeds of 80 tokens per second and offers twice the token efficiency of other leading models, leading to quicker, more intelligent results at lower costs. Grok 4.5 is now the default model in Grok Build and is priced at $2 per million input tokens and $6 per million output tokens. This means you can expect more efficient and cost-effective performance for a wide range of tasks.