DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

215 points · 78 comments on HN · read original →

Points and comments are a snapshot, not live.

DeepSeek V4 Flash 0731 offers strong performance at roughly one-third the cost of OpenAI Luna Max effort.

The article is a 404 page; no analysis was available.

What commenters are saying

Several commenters compare DeepSeek V4 Flash to OpenAI Luna on cost and speed. One notes DeepSeek is about 2-3x cheaper than Luna Max effort but 2-5x slower. Others praise its cost-effectiveness for coding tasks, with one developer saying it allows a generous free tier for their app. A few discuss censorship, with claims that Western models censor differently than Chinese ones. Some express anticipation for an updated DeepSeek V4 Pro with improved performance. A comment provides a speed benchmark of 93 TPS on OpenRouter.