InfraAug 19, 2026 · 09:482 sources
Perplexity and Baseten Host DeepSeek V4 Pro in US, Cutting Frontier Costs 62%
From huggingnews.com
Perplexity and Baseten have hosted the 0813 version of DeepSeek V4 Pro in the U.S., achieving 147 tokens per second in tests by Artificial Analysis, the highest speed among current inference providers. Perplexity integrated the model with a $0.75 per task cost and a WANDR score of 0.359, reporting it is 62% cheaper than the next closest option on the cost-performance frontier. Baseten is continuing to optimize throughput and latency for the U.S.-hosted model.
Read the sourceSign in to suggest
Sources
+ 1 more
- SUPPORT@baseten“fastest inference provider on Artificial Analysis for DeepSeek V4 Pro 0813, at 147 TPS”x.com
#deepseek#inference#llm-pricingMarkdown
Content history (2)
- 2026-10-16 · Summary · vi · Replaced stored textTrong các thử nghiệm của Artificial Analysis, phiên bản 0813 của DeepSeek V4 Pro đạt tốc …Phiên bản 0813 của DeepSeek V4 Pro đạt 147 token/giây trong kiểm tra bởi Artificial Analy…
- 2026-10-16 · Title · vi · Replaced stored textPerplexity và Baseten đưa DeepSeek V4 Pro lên hạ tầng tại Mỹ, giúp giảm 62% chi phíPerplexity và Baseten triển khai DeepSeek V4 Pro tại Mỹ, giảm chi phí tiền phong 62%
One ranked edition, every morning
The stories that matter today. Free, cancel anytime.
Telegram @aidr_today