Elon Musk’s SpaceXAI has released Grok 4.5, its first public model since the SpaceX-xAI merger closed in February. The model targets coders, engineers, and knowledge workers, with a pricing strategy that undercuts competitors. Grok 4.5 costs $2 per million input tokens and $6 per million output tokens, compared to Claude Opus 4.8’s $5 input and $25 output, and GPT 5.6’s $5 input and $30 output.
Benchmark Results
Grok 4.5’s performance is mixed, according to four benchmark results published by SpaceXAI. On DeepSWE 1.1, the model scored 53%, behind Claude Opus 4.8’s 59% and GPT 5.5’s 67%. However, on SWE Bench Pro, Grok 4.5 posted 64.7%, beating GPT 5.5’s 58.6%.
Training and Efficiency
SpaceXAI trained Grok 4.5 on tens of thousands of Nvidia GB300 GPUs, resulting in a model that is competitive, but not top-performing. However, the model’s efficiency is a key selling point, using an average of 15,954 output tokens to complete each job on SWE Bench Pro tasks, compared to Opus 4.8’s 67,020 tokens.
Availability and Pricing
Grok 4.5 is available via API, on Hermes and Grok build, with half a million tokens of context. European users will have to wait until mid-July to access the model. With its lower pricing and efficient usage, Grok 4.5 may appeal to teams running high-volume coding tasks, despite its lower performance on quality benchmarks.
Based on reporting from crypto.news.