SpaceXAI Unleashes Grok 4.5: An "Opus-Class" Model with Aggressive Pricing to Disrupt AI Landscape

SpaceXAI's Strategic Entry into the AI Frontier

SpaceXAI, led by Elon Musk, officially released its latest artificial intelligence model, Grok 4.5, on Wednesday, July 8, 2026. This launch marks a significant move for the company, positioning Grok 4.5 as a direct competitor to top-tier models from rivals like Anthropic and OpenAI. Musk himself described Grok 4.5 as an "Opus-class model," emphasizing its competitive performance, superior speed, enhanced token efficiency, and significantly lower cost.

The development of Grok 4.5 is a direct result of SpaceXAI's recent $60 billion acquisition of the AI coding startup Cursor. This integration allowed for the model to be trained specifically for coding and autonomous agents, making it the first tangible product of Musk's vertically integrated AI strategy. The model runs on xAI's new 1.5-trillion-parameter V9 foundation model, a substantial architectural upgrade from previous versions.

Aggressive Pricing and Token Efficiency

One of the most compelling aspects of Grok 4.5 is its aggressive pricing strategy, which aims to significantly undercut its competitors. SpaceXAI has priced Grok 4.5 at $2 per million input tokens and $6 per million output tokens. This pricing is notably lower than premium tiers offered by rivals; for instance, Anthropic's Opus 4.8 runs at $5 and $25 per million tokens, respectively, while OpenAI's GPT-5.6 Luna is priced at $1 and $6.

Beyond just lower per-token costs, Grok 4.5 boasts impressive token efficiency, reportedly using half as many tokens per task as comparable models and delivering higher throughput. Artificial Analysis, a benchmarking firm, measured Grok 4.5 at $0.49 per completed task, noting it was "nearly 90% cheaper than the models ahead of it on our leaderboard" and placed it "clearly on the Pareto frontier for performance versus cost." This efficiency is particularly crucial for agentic workloads, which consume tokens voraciously.

Designed for Coding and Agentic Workloads

Grok 4.5 is specifically engineered for coding and agentic tasks, making it a powerful tool for developers and enterprises. It was trained on extensive datasets covering coding, science, engineering, and mathematics, with supplemental training incorporating Cursor's developer data. The model excels at handling difficult, long-running tasks, including complex software engineering problems.

Key capabilities and features of Grok 4.5 include:

• Coding Proficiency: Highly capable in various programming languages, including Rust and C/C++, and adept at end-to-end app building from a single prompt.
• Agentic Workflows: Designed for autonomous multi-step task completion across various domains.
• Office Integration: Offers plugins for Microsoft Word, PowerPoint, and Excel, enabling it to build complex Excel models with web research, design diagrams, and write structured prose.
• Broad Application: Beyond coding, it targets legal and financial work, and includes cybersecurity features.

Performance Benchmarks and Availability

While Elon Musk has stated that Grok 4.5 is "roughly comparable to Opus 4.7, but much faster," independent evaluations present a mixed picture. On the Artificial Analysis GDPval-AA v2 index for real-world agentic knowledge work, Grok 4.5 ranked fourth, behind the latest Claude releases from Anthropic. In coding benchmarks, Grok 4.5 scored 83.3% on Terminal Bench 2.1, nearly matching GPT 5.5 (83.4%) but trailing Anthropic's Fable 5 (84.3%). On DeepSWE 1.1, which measures the ability to resolve real GitHub issues, Grok 4.5 achieved 53%, behind OpenAI's GPT-5.5 at 67% and Fable 5 at 70%.

Despite some benchmark gaps, the model's cost-efficiency could be a significant differentiator. Grok 4.5 is currently available in the U.S. through Grok Build, Cursor, and the SpaceXAI console, with a wider public release expected to expand to the European Union by mid-July. It is also accessible via the xAI API and supported by model gateways including OpenRouter, Vercel, Cloudflare, Snowflake, and Databricks Mosaic.