Model3h ago
xAI Launches Budget-Friendly Grok 4.5
Grok 4.5 costs $2 per million input tokens. Trained on Nvidia GB300 GPUs, the model trails GPT-5.5 in coding benchmarks but requires 4.2 times fewer tokens than Opus 4.8. This pricing strategy prioritizes cost-efficiency over raw performance. Developers can now trade slight benchmark deficits for significantly lower operational overhead.