StorySoftware
xAI releases Grok 4.5
SpaceX's xAI released Grok 4.5, aimed at coding, agent tasks and office work, trained on tens of thousands of NVIDIA GB300 GPUs. It scored 62% on the DeepSWE 1.0 software benchmark, runs at about 80 tokens per second and costs $2 per million input tokens and $6 per million output tokens via Grok Build, Cursor and the API.
- Trained on tens of thousands of NVIDIA GB300 GPUs; doubles token efficiency of comparable leading models.
- DeepSWE 1.0: 62.0% (Fable 66.1%, GPT 5.5 64.31%, Opus 4.8 55.75%).
- Speed: ~80 tokens per second; pricing $2/$6 per million input/output tokens via Grok Build, Cursor and API.