xAI just dropped Grok 4.5 on July 8, and its first order of business was climbing to the top of the SWE Marathon benchmark with a 29.0% resolution rate. That puts it ahead of Anthropic’s Claude Opus 4.8 at 26.0% and Fable at 24.0%, making it the current king of long-horizon software engineering tasks.
What Grok 4.5 actually brings to the table
The model was trained on tens of thousands of NVIDIA GB300 GPUs, with reinforcement learning specifically tuned for software engineering tasks. Inference speed clocks in at roughly 80 transactions per second. xAI claims up to 4.2 times better token efficiency compared to leading competitors in specific applications.
At $2 per million input tokens and $6 per million output tokens, xAI is positioning Grok 4.5 as an “Opus-class” model at a distinctly non-Opus price point. The model launched via the Grok app, xAI console, and API, with EU availability expected by mid-July. Beyond SWE Marathon, Grok 4.5 posted strong results across DeepSWE and Terminal Bench 2.1. One caveat worth noting: there’s been no independent third-party verification of the SWE Marathon results yet.
Why crypto markets should pay attention to the AI coding wars
