SpaceXAI highlighted benchmark results comparing Grok 4.5 with other leading AI models across software engineering evaluations. The company said the model achieved competitive results on DeepSWE, Terminal Bench, and SWE Bench Pro, while also ranking first on Harvey’s Legal Agent Benchmark for office-related tasks.
The company said Grok 4.5 was trained across tens of thousands of NVIDIA GB300 GPUs using data filtering, quality scoring, deduplication, and domain-specific curation techniques. Reinforcement learning focused on multi-step software engineering and technical tasks, with training spanning hundreds of thousands of evaluations using automated and model-based grading.
SpaceXAI also emphasized the model’s ability to generate complete applications from a single prompt. It showcased examples ranging from a Three.js solar system simulation to office productivity tasks involving Excel, PowerPoint, and Word. According to the company, Grok 4.5 can build complex spreadsheets that incorporate web research and multi-sheet formulas, while also creating presentation decks and structured documents using native Microsoft Office features.
The company said Grok 4.5 is served at speeds of 80 tokens per second and delivers approximately twice the token efficiency of comparable leading models on the same tasks. In its published results, SpaceXAI reported that Grok 4.5 completed SWE Bench Pro tasks using an average of 15,954 output tokens, compared with 67,020 for Opus 4.8 (max).
SpaceXAI priced Grok 4.5 at $2 per million input tokens and $6 per million output tokens, saying its pricing and token efficiency allow the model to deliver higher intelligence per unit of time and cost.
In a post on X, founder Elon Musk described the model as competitive with Anthropic’s Opus series. “Based on strong positive feedback from customers in our beta test program, @SpaceXAI will make Grok 4.5 available to the public tomorrow. It is an Opus-class model, but faster, more token-efficient and lower cost,” Musk wrote.
He later added, “Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster. The combination of capability, faster speed and lower cost is what makes it competitive.”
Grok 4.5 is available beginning today in Grok Build, Cursor on all plans, and through the SpaceXAI console. SpaceXAI noted that the model is not yet available in the European Union through its products or API console, with EU availability expected in mid-July.
This analysis is based on reporting from SpaceXAI.
Images courtesy of SpaceXAI.
This article was generated with AI assistance and reviewed for accuracy and quality.
