SpaceXAI just launched Grok 4.5, a coding-focused AI model trained with Cursor data. It promises lower costs and faster speeds for autonomous agent tasks.
SpaceXAI just dropped Grok 4.5, its most capable model yet. Designed specifically for coding and autonomous “agentic” tasks, the company positions this launch as a direct challenge to industry leaders like Anthropic’s Claude Opus.
The model’s secret sauce?
Training data from Cursor, the AI-powered code editor that SpaceXAI acquired last month for $60 billion. By combining that real-world developer data with a massive 1.5-trillion-parameter foundation, the team built a model that supposedly solves complex engineering tasks with significantly less “token burn” than its rivals.
Elon Musk claims Grok 4.5 matches the intelligence of Claude Opus but delivers results faster and at a much lower cost. Pricing reflects that aggressive strategy: users pay $2 per million input tokens and $6 per million output tokens.
While benchmarks show mixed results compared to other frontier models, the efficiency gain is undeniable. SpaceXAI reports that Grok 4.5 uses roughly 4 times fewer output tokens than leading models on technical benchmarks, saving developers both time and money during heavy agentic workloads.
You can access Grok 4.5 right now through the SpaceXAI console, Grok Build, and the Cursor editor. European users, however, have to wait a little longer; SpaceXAI expects to roll out access there later this month.
With models like Grok 4.5 moving toward cheaper, more efficient agentic coding, will autonomous programming replace human developers in their own workflow, or will the industry prefer to keep a hand on the wheel?
