Open-weight models now account for 56% of total token volume flowing through Vercel AI Gateway as of August 2026, up from just 11% in April, a fivefold increase in four months. However, this trend masks a paradox: Anthropic, maker of closed Claude models, still captures 61-65% of total gateway spend despite only 30% of tokens processed, thanks to price premiums up to 4.4 times the average rate. DeepSeek leads in volume share with approximately 25%, while the average price per token fell 23.2% in August alone, a drop exceeding 50% over five months. Open-weight models now run high-volume workloads at roughly one-seventh the cost of frontier closed models.
Source: Read the original article
