DeepSeek
DeepSeek Tops AI Tiger Compute
Among AI Tigers, DeepSeek's V4-Pro leads documented open-weight training compute at 9.7e24 FLOP. Alibaba's Qwen2.5 Instruct follows at 7.9e24 FLOP.
DeepSeek
Among AI Tigers, DeepSeek's V4-Pro leads documented open-weight training compute at 9.7e24 FLOP. Alibaba's Qwen2.5 Instruct follows at 7.9e24 FLOP.
DeepSeek
DeepSeek V4 Pro scored 51.5 on Artificial Analysis at $0.87 per million output tokens on OpenRouter. Claude Opus 4.8 scored 61.4 but cost $25 – 29x more.
DeepSeek
In May 2026, among AI Tigers, DeepSeek V4 Pro had 5.81M downloads – second-most on Hugging Face. Qwen3.6-35B-A3B led by 56,620 downloads.
DeepSeek
In May 2026, among AI Tigers, DeepSeek models processed 19.51T tokens – the most on OpenRouter. DeepSeek V4 took it from third in April to first in May.
DeepSeek
In May 2026, DeepSeek models on OpenRouter processed 19.51 trillion tokens. DeepSeek tokens processed increased 187.7% month-over-month.