DeepSeek’s V4-Flash is the cheapest well-known AI model to run, research firm finds

It now costs about three cents to push one of the world’s better-known AI models through a full benchmark suite. That is the figure from Artificial Analysis, which found that a version of DeepSeek’s flagship model, V4-Flash, is by far the cheapest well-known model to run.

The independent research firm measured the cost of completing its Intelligence Index test battery on each model. V4-Flash came in at roughly three cents, and the nearest comparisons were not close: Moonshot’s Kimi K3 cost 86 cents, OpenAI’s GPT-5.6 Sol $1.86, and Anthropic’s Claude Fable 5 $3.15.

On published pricing, DeepSeek charges $0.14 per million input tokens and $0.28 per million output tokens for the model. Those are the sort of numbers that quietly redefine what ‘expensive’ means at the frontier.

V4-Flash is the lighter half of the pair DeepSeek shipped when it returned with V4-Pro and V4-Flash, with the heavier V4-Pro aimed at harder reasoning. Flash is built for volume: fast, cheap, and good enough for a large share of everyday work.

The catch is capability, and the numbers are honest about it. V4-Flash scored 50 out of 100 on the Intelligence Index, level with Google’s Gemini 3.6 Flash and just behind Meta’s Muse Spark 1.1 and Z.ai’s GLM-5.2, both on 51.

The frontier still sits clearly ahead. Kimi K3 scored 57, while Claude Opus 5, Claude Fable 5, and GPT-5.6 landed roughly nine points higher again, a reminder that cheapest and best remain different questions.

The release lands in the middle of an AI price war that DeepSeek has done more than anyone to start. The company made a 75% discount permanent earlier this year, and rivals have been cutting in response.

Those rivals have been moving the same way. OpenAI trimmed GPT-5.6 pricing sharply, and the general drift of the market has been down, and fast, on a curve that looks less like software margins and more like a commodity.

DeepSeek has the balance sheet to keep pushing. The company recently closed its first outside funding, a round of more than $7bn, which buys room to subsidise aggressive pricing while it takes share.

Flash-class models are aimed at the high-volume end of the market. That means the chatbots, coding assistants, and back-office automation where requests run into the millions and every fraction of a cent compounds into a real bill.

DeepSeek’s edge is as much engineering as pricing. The company has leaned on efficient training and inference to hold costs down, which is what lets it charge so little without, it says, simply setting money on fire.

That trend carries consequences beyond a cheaper API bill. Analysts have argued that relentless discounting from Chinese labs puts the eventual OpenAI and Anthropic IPOs under pressure, since premium pricing is hard to defend when a rival is tens of times cheaper per task.

Not everyone is convinced the quality gap still matters. Zack Kass, OpenAI’s former head of go-to-market, has framed the moment as one of ‘diminishing model returns’, arguing that once models are close enough, the next one barely moves the needle and price does the deciding.

Chinese labs have been setting that pace. Moonshot’s Kimi K3 spooked markets on release, and the broader worry is that a wave of cheap, open-weight models erodes the economics Western AI valuations quietly assume.

Benchmarks are an imperfect proxy, and cost per test turns on how efficiently a model spends tokens as much as on its sticker price. Even so, the direction is not in doubt, and Artificial Analysis has put hard figures on what developers have felt for months.

For buyers, the sum is getting simpler. If a model that costs three cents to run can do most of the job, the burden shifts onto the expensive models to prove what those extra nine points on a benchmark are really worth.