SpaceXAI's Grok 4.5 Disrupts AI Market with Aggressive $2/M Token Pricing

SpaceXAI's new Grok 4.5 model, launched yesterday, immediately disrupts the competitive AI landscape with an aggressive $2 per million input token pricing strategy, significantly undercutting rivals while delivering high-performance capabilities. This Cursor-trained model, developed jointly with the recently acquired AI coding editor, targets coding, complex agentic tasks, and extensive knowledge work, promising a new era of efficiency and accessibility for developers and enterprises alike.

Grok 4.5, now available in Grok Build, Cursor, and via API, boasts a substantial 500,000-token context window. It is engineered for real-world excellence, leveraging training across tens of thousands of NVIDIA GB300 GPUs and trillions of tokens from Cursor data, capturing intricate developer interactions with codebases and software tools. The model delivers a blistering 80 tokens per second (TPS) and claims twice the token efficiency of leading models, resolving tasks in under half the steps. While some broader benchmarks show competitors leading in specific categories like DeepSWE 1.1, Grok 4.5 holds its own, nearly matching GPT 5.5 and Fable 5 on Terminal Bench 2.1. Crucially, it ranks #1 on Harvey's Legal Agent Benchmark, indicating strong proficiency in complex legal reasoning and document generation.

This release signifies SpaceXAI's serious intent to capture a substantial share of the enterprise AI market, moving beyond general-purpose chatbots. The integrated training with Cursor showcases a strategic vertical integration, aiming to create highly specialized, developer-native AI tools that are both powerful and practical. At $2 per million input tokens and $6 per million output tokens, Grok 4.5's pricing dramatically undercuts models like Anthropic's Opus 4.8 ($5/$25) and Fable 5 ($10/$50), positioning it as a compelling cost-performance leader. This aggressive move could force a broader industry recalibration of AI model economics, democratizing access to powerful agentic capabilities for businesses previously deterred by high operational costs and setting a new bar for AI utility.