Grok 4.5 Tops Real-World Invoice Processing, Beating Gemini and GPT

Ramp tested AI models on 150k real business invoices. Grok 4.5 achieved the highest perfect-extraction rate, outperforming Gemini Flash 3.6, GPT 5.6 Terra, and Sonnet 5. The task requires long-context reasoning over 100K+ tokens.

• 长上下文
• AI比较
• Grok 4.5
• Ramp
• 发票处理