SpaceX Launches Grok 4.5 AI Model Focused on Coding and Agents

Elon Musk's SpaceX launched Grok 4.5 on Wednesday, marking the company's first artificial intelligence model trained specifically for coding tasks and autonomous agents. This release also represents the first product stemming from SpaceX's recently completed $60 billion acquisition of AI coding startup Cursor.

The launch serves as a critical test for Musk's expanding AI endeavors, emphasizing a strategy that prioritizes speed, cost-effectiveness, and practical application over leaderboard performance for developers.

SpaceX is positioning Grok 4.5 not as the most capable model, but as the most economical. The company states the model uses fewer tokens per task than comparable models, offers higher throughput, and is priced at less than half that of competitors. Input tokens cost $2 per million, and output tokens are $6 per million, significantly undercutting offerings from rivals such as Anthropic and OpenAI.

Independent analyses suggest Grok 4.5 is competitive in real-world agentic tasks, ranking fourth in performance according to benchmarks, but its cost efficiency is a standout feature. Analysts at Stone & Company noted it is nearly 90% cheaper per completed task than leading models, making it a compelling option for enterprises deploying autonomous agents at scale for software development.

The development of Grok 4.5 is directly tied to the strategic rationale behind the $60 billion Cursor acquisition. SpaceX integrated Cursor's AI-first code editor, which generates extensive interaction data from expert engineers, into Grok's training process. This synergy also provides Cursor access to SpaceX's substantial computing infrastructure.