Cursor has rolled out support for Grok 4.5, SpaceXAI’s latest large language model optimized for coding, agentic tasks, and complex knowledge work. Grok 4.5 is now available as a core engine inside Cursor’s AI coding environment, giving developers access to a high‑capacity model directly within their editor and workflows.
Grok 4.5 was trained in close collaboration with Cursor on real developer usage, with SpaceXAI running training across tens of thousands of Nvidia GB300 GPUs and emphasizing aggressive data filtering, deduplication, and domain‑focused curation. The model is described as “Opus‑class” in terms of intelligence—comparable to Anthropic’s Claude Opus 4.8—while being faster, more token‑efficient, and lower cost.
The model supports both text and image input and is built for long‑context coding sessions, with a context window of up to 500,000 tokens. On coding benchmarks such as DeepSWE and SWE Bench Pro, Grok 4.5 matches or outperforms top rival models while generating significantly fewer output tokens per task, improving speed and cost efficiency for iterative development work.
Through Cursor and Grok Build, Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens, undercutting many models with similar performance profiles. Developers can access Grok 4.5 across all Cursor plans and via the SpaceXAI console using an API key, with EU availability expected later in July 2026.
