Grok 4.5 Tested: Bigger Context, Cheaper and Faster

socialwithaayan · x · 2026-07-18

The author didn't expect to be so impressed by Grok 4.5, but testing it changed their mind.

Key highlights include:

A context window of 500K tokens, capable of ingesting entire project files at once while retaining details;
Impressive speed, with responses much faster than expected;
Costs roughly 60% cheaper per task compared to Claude Opus 4.8;
Ranks 4th on the Artificial Analysis leaderboard;
Reportedly trained alongside Cursor, making it better suited for multi-step tasks in real code repositories;
Direct integration with X's real-time data.

The author concludes that it’s not just a "cheap alternative," but a smarter default choice for many non-throwaway tasks.

More from Models