xAI launches Grok 4.5 with 1.5T parameters at 40% of competitor pricing
xAI launched Grok 4.5 on July 9 with 1.5T parameters, priced at $2 input and $6 output per 1M tokens, achieving 80 TPS inference speed.
xAI (now merged into SpaceXAI) officially released Grok 4.5 on July 9 Beijing time. The model is based on xAI's self-developed V9 base architecture with 1.5T parameters, three times the size of the previous generation Grok 4.3.
Grok 4.5 was jointly trained with Cursor, integrating trillions of developer interaction data points from the Cursor platform, including debug traces, multi-file diffs, and user correction patterns. Training was completed on tens of thousands of NVIDIA GB300 GPUs.
Pricing for Grok 4.5 is $2 input and $6 output per 1M tokens, more than 60% cheaper than Claude Opus 4.8 and GPT-5.5. Inference speed reaches 80 tokens/s. On SWE-Bench Pro, Grok 4.5 consumes less than a quarter of the tokens that Opus 4.8 Max needs to complete the same tasks.
Musk stated that xAI will maintain a monthly iteration cadence going forward, with plans to release a 1M token context version soon and a 2T parameter version by month end. Grok 4.5 has been integrated into Grok Build, all Cursor plans, and the xAI developer console.