The company xAI announced that it will skip the Grok 3.5 version and move directly to the release of Grok 4. The new model will appear after July 4, but the exact date is unknown. The main focus will be on improving programming capabilities. There will even be a special version “Grok 4 Code” aimed at developers. The novelty features an integrated code editor in the style of VSCode, allowing the model to write, modify, and debug code directly in the Grok web interface.
Grok 4 will support text and visual queries. The image generation feature is still being refined. The model will have a context window of 130,000 tokens, which is less than some competitors, but this may speed up work and enhance user convenience. It provides function calls, structured responses, and advanced logical analysis, making Grok 4 a versatile tool for working with text, math, and code.
Grok 4 Code will be an assistant in environments like Cursor, and will also offer deep integration with other code editors. The model will work through API and web interface, allowing its use in both professional and consumer tasks. Updates have already appeared in the xAI developer console, mentioning two versions — Grok 4 and Grok 4 Code.
Initial tests of Grok 4 indicate a significant increase in performance, especially in logic and text comprehension tasks. For example, in the Humanity last-exam test, the model scored 35 percent, and with additional calculations — 45, exceeding the results of competitors like Gemini 2.5 Pro, o3 Pro, and Claude 4 Opus.
