Grok 4 Coding

Leads raw SWE-bench at 75%. Purpose-built for software engineering tasks.

Grok 4 Coding is xAI’s specialized software engineering model, leading raw SWE-bench scores at 75%. Unlike general-purpose models that were tuned for coding secondarily, Grok 4 Coding was trained from the ground up with software engineering as the primary objective. It excels at multi-file refactoring, bug detection across large codebases, and generating functional code from ambiguous natural language specifications.

Grok 4 Coding

Made by

xAI

Released

2026

Context Window

Large

Pricing

~$2 / $15 per 1M tokens

Provider

xAI