DeepSeek's V4-Flash is the cheapest major AI model to run

What happened: DeepSeek released V4-Flash on 31 July. Research firm Artificial Analysis found it is by far the cheapest well-known AI model to run, according to Reuters via Yahoo Finance.

The numbers: V4-Flash costs 3 cents per test on average, against 86 cents for Moonshot's Kimi K3, $1.86 for OpenAI's GPT-5.6 Sol, and $3.15 for Anthropic's Claude Fable 5. That makes it more than 100 times cheaper than Fable 5. List price sits at $0.14 per million input tokens and $0.28 per million output tokens.

The context: Artificial Analysis measures cost per test, not just list price, because a model that looks cheap per token can still end up pricier if it needs many more steps to reach the right answer, according to The Next Web.

The catch: Cheap doesn't mean best. V4-Flash scores 50 out of 100 on the Artificial Analysis Intelligence Index, level with Google's Gemini 3.6 Flash and a point behind Meta's Muse Spark 1.1 and Z.ai's GLM-5.2. Kimi K3 scores 57. Claude Opus 5, Fable 5 and GPT-5.6 all lead by at least nine points.

The details: V4-Flash is a mixture-of-experts model with about 284 billion total parameters, only around 13 billion active per request. It uses hybrid attention and supports a 1 million token context window.

Why it matters: For routine, high-volume tasks, price is becoming the deciding factor. Chinese labs keep pushing that price towards zero.

Who's affected: DeepSeek is reportedly preparing for a possible IPO. Its R1 model triggered a global tech sell-off in early 2025.

What's next: Buyers weighing routine AI workloads now have a far cheaper option, even as top-tier accuracy still sits with pricier models like Fable 5 and GPT-5.6.