Research: DeepSeek's Latest V4-Flash Is Currently the Lowest-Cost Model to Run
I'm LongbridgeAI, I can summarize articles.Research firm Artificial Analysis reports that DeepSeek's V4-Flash AI model is the lowest-cost to run, costing over 100 times less than Anthropic's Claude Fable 5. Priced at $0.14 per million input tokens, it significantly undercuts competitors like OpenAI and Moonshot AI. Despite its cost efficiency, V4-Flash scored 50/100 in intelligence ratings, trailing behind models from Meta, Z.AI, and Moonshot.
Research firm Artificial Analysis data showed that DeepSeek's flagship artificial intelligence model V4-Flash has operating costs far below other leading global models in benchmark tests, costing more than 100 times less than Anthropic's Claude Fable 5.
V4-Flash charges USD0.14 per million input tokens and USD0.28 per million output tokens, with the average estimated cost per test at USD0.03. In comparison, the cost of Chinese competitor Moonshot AI's Kimi K3 is USD0.86, while OpenAI's GPT-5.6 Sol costs USD1.86 and Claude Fable 5 costs USD3.15.
Artificial Analysis also said that the V4-Flash model scored 50 points out of 100 in intelligence quotient ratings, the same score as Google's Gemini 3.6 Flash, and 1 point lower than Meta Platforms, Inc. (META.US) 's Muse Spark 1.1 and Z.AI (02513.HK) -60.500 (-6.127%) Short selling $358.48M; Ratio 9.811% 's GLM-5.2. However, Moonshot AI's Kimi K3 scored 57 points, while Anthropic's Claude Opus 5, Fable 5 and OpenAI's GPT-5.6 scored 9 points or more higher.(mn/a)(HK stocks quote is delayed for at least 15 mins.Short Selling Data as at 2026-08-03 12:25.) (Real-time Streaming US Stocks Quote; Except All OTC quotes are at least 15 minutes delayed.)