SpaceX's Grok 4.5 launches at half the price of rivals — here's why that could rattle Anthropic and OpenAI

2026-07-08 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

SpaceX launched Grok 4.5 on Wednesday, July 8, 2026, its first artificial intelligence model specifically trained for coding and autonomous agents. This release is the initial tangible product of SpaceX's recent \$60 billion acquisition of AI coding startup Cursor. Grok 4.5 is positioned as a cost-effective alternative to frontier models, priced at \$2 per million input tokens and \$6 per million output tokens, which is less than half the cost of rivals like Anthropic's Claude Opus. While not claiming to be the smartest, internal assessments and independent evaluations by Artificial Analysis suggest it is competitive, achieving an Elo score of 1543 on the GDPval-AA v2 index and costing \$0.49 per completed task, nearly 90% cheaper than top models. The model's training utilized Cursor's interaction data and SpaceX's Colossus supercomputer, aiming for real-world usefulness in complex codebases.

Key takeaway

For Directors of AI/ML evaluating coding models, Grok 4.5's aggressive pricing and competitive performance demand your attention. If your organization deploys agentic workloads, its 90% lower cost per completed task could significantly reduce operational expenses, even if raw capability is slightly below top-tier rivals. You should pilot Grok 4.5 for specific engineering tasks to assess its real-world utility and "vibes" within your team, potentially shifting your procurement strategy towards cost-efficiency.

Key insights

SpaceX's Grok 4.5 disrupts the AI coding market by prioritizing cost-efficiency and speed over raw benchmark dominance.

Principles

• Developers prioritize speed, cost, and utility over benchmark scores.
• Vertical integration of AI stack creates competitive advantage.
• Cost-per-task can outweigh marginal capability differences.

Method

Grok 4.5 was trained using high-quality interaction data from Cursor's AI-first code editor, leveraging SpaceX's Colossus supercomputer with 200,000 Nvidia GPUs.

In practice

• Deploy cheaper models for agentic workloads to reduce token consumption.
• Evaluate AI models on real-world usefulness and cost-per-task.
• Consider integrated AI stacks for end-to-end control and efficiency.

Topics

• Grok 4.5
• AI Coding Models
• Cost-Performance Frontier
• Vertical Integration
• Autonomous Agents
• Enterprise AI

Best for: CTO, VP of Engineering/Data, Machine Learning Engineer, AI Engineer, Director of AI/ML, Investor

Related on AIssential

• Codex is gaining steam — AI platforms are evolving to be more accessible to non-technical users and offer competitive performance-to-cost ratios.
• Anthropic-SpaceXai's 300MW/$5B/yr deal for Colossus I, ARR growth is 8000% annualized — Compute capacity and agent system ergonomics are now as critical as raw model performance for frontier AI labs.
• Elon, stop trying to make Grok happen — Grok's poor performance and controversial outputs undermine its market viability and SpaceX's IPO strategy.
• xAI launches Grok Build coding agent for developers — xAI launched Grok Build, a coding agent, amid competitive pressures and internal restructuring following its acquisition by SpaceX.
• Grok 4.2: Powerful and Fast Coding Model! Most Powerful & Intelligent Model We've Seen? (Early Test) — Grok 4.2 is an upcoming model with enhanced multimodal capabilities and efficiency, despite some early coding limitations.
• OpenAI's GPT-5.5 is here, and it's no potato: narrowly beats Anthropic's Claude Mythos Preview on Terminal-Bench 2.0 — GPT-5.5 redefines AI agency, excelling in complex, multi-part tasks with less human guidance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.