Grok 4.5 release
Grok 4.5 is being released by SpaceXAI and is positioned as a cost-effective model aimed at coders, engineers, and knowledge workers, emphasizing cost efficiency. Pricing is set at $2 per million input tokens and $6 per million output tokens. The company frames Grok 4.5 as cheap rather than the best model, trading speed and cost efficiencies against raw capability.
Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens and is positioned by the company as a low-cost offering optimized for speed and token efficiency. By comparison, Claude Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens, while GPT 5.6 Sol is priced at $5 per million input tokens and $30 per million output tokens. The company frames Grok 4.5 as cheap rather than the best model, presenting a tradeoff of speed and cost efficiencies against raw capability.
The better argument isn’t raw performance; it’s efficiency math.
Grok 4.5 was trained in collaboration with Cursor AI during its development cycle, according to the provided details. The training process used tens of thousands of Nvidia GB300 GPUs allocated for the model’s training runs, as reported in the same material. These runs occurred inside Colossus, the Memphis supercomputer, which is described as having total capacity across more than 200,000 GPUs.
SpaceXAI’s merger with xAI closed in February, a recent corporate change noted alongside Grok 4.5. Separately, SpaceX has a pending $60 billion deal to acquire Cursor AI, a transaction mentioned in the source material. The model is targeted at coders, engineers, and knowledge workers.
Grok 4.5 posted a 64.7% resolution rate on the SWE Bench Pro benchmark. By comparison, GPT 5.5 posted 58.6%, Opus 4.8 posted 69.2%, and Fable 5 posted 80.4%. Grok 4.5’s score is higher than GPT 5.5’s and lower than Opus 4.8’s and Fable 5’s, placing it between those models on this metric. The company frames Grok 4.5 as a model optimized for lower cost and greater speed rather than peak performance, which aligns with its SWE Bench Pro result not being the top score. These figures are the SWE Bench Pro resolution rates provided for each model.
Grok 4.5 is presented by SpaceXAI as a cost-efficient AI model designed for coders, engineers, and knowledge workers, aimed at delivering lower operating costs and token efficiency for everyday development and knowledge‑work tasks.
The company frames Grok 4.5 as prioritizing speed and reduced cost over maximizing raw performance, describing this as a deliberate tradeoff between efficiency and capability rather than an attempt to be the highest‑performing model.
