Deepseek V4 Flash 0731 Scores 50 On The Artificial Analysis Intelligence Index 10 Points Above Previous Deepseek V4 Flash
Source
artificialanalysis.ai
Author
Artificial Analysis articles
Date
Why it matters
DeepSeek V4 Flash 0731 scores 50 at roughly 60% lower cost per task than a comparable OpenAI model, helped by a ~98% cache hit discount, with weights expected soon.
Key takeaways · AI-distilled
The July 31 evalA repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.Full definition → puts DeepSeek V4 Flash 0731 at 50 on the Intelligence Index, ten points above its predecessor, without changing its stated parameter count or tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → prices.
DeepSeek V4 Flash 0731's AA-Omniscience gain came from fewer hallucinations, with accuracy unchanged. The improved index does not show broader factual knowledge.
DeepSeek V4 Flash 0731's reported cost advantage depends partly on its first-party API cache discount of about 98%. Different providers or workloads with fewer cache hits may not realize the same savings.
Terms in this piece · Glossary
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
