Journalism of Courage
DeepSeek launches V4 Flash, one of the cheapest models to run: Is AI now a commodity?
DeepSeek is the latest Chinese tech startup bearing down on the US market, with a full-scale price war unfolding across the AI industry over the past few months.
Tech industry leaders such as Microsoft CEO Satya Nadella have repeatedly argued that AI will emerge as a commodity. (File photo)
Amid the rapidly intensifying AI race between the US and China, Chinese AI startup DeepSeek has launched a powerful new code-generation AI model called V4 Flash.
V4 Flash is an open-weight AI model, available under the permissive MIT License which allows for free commercial use, modification, and local deployment. Its model weights are hosted on platforms like Hugging Face. The model’s underlying mixture-of-experts (MoE) architecture means that out of its 284 billion total parameter count, 13 billion parameters are activated for efficient reasoning across a one million-token context window.
However, V4 Flash’s biggest draw is likely its usage costs. DeepSeek charges about 28 cents for the same amount of coding output that costs $25 on Anthropic’s Claude Opus 4.8 model, making the Chinese open-weight model about 99 per cent cheaper to use.
V4 Flash is also said to come close to the level of Opus 4.8 in terms of performance. When evaluated on Arena.ai, a crowdsourced leaderboard for front-end coding, V4 Flash beat Opus 4.8, delivering the best performance for its price among any model in its class.
Widely known for triggering a market meltdown last January by releasing a cutting-edge AI model with far fewer resources than its US rivals, DeepSeek is the latest Chinese AI startup bearing down on the US market, with a full-scale price war unfolding across the AI industry over the past few months.
While tech giants such as Google and Meta have continued to pour hundreds of billions of dollars into the computing infrastructure powering the AI revolution, industry trends suggest that intelligence is getting cheaper by the week. The debut of V4 Flash, which is considered to be one of the cheapest models to run, is also the latest indicator that AI is rapidly becoming a commodity.
Essentially, it means that the shrinking performance gap between top-tier models will lead to many AI applications that do not depend on just a single model provider. It gives buyers more leverage to shop on price.
Story continues below this ad
Tech industry leaders such as Microsoft CEO Satya Nadella have repeatedly argued that AI will emerge as a commodity. In a recent essay posted on X, Nadella also warned against relying on just a handful of AI model providers, and urged companies to build and retain their own AI capabilities alongside human expertise rather than ceding organisational knowledge to a small number of AI models.
However, falling prices may not necessarily spell doom for frontier AI labs such as OpenAI and Anthropic as it could end up unleashing vastly more demand. “We will have so much usage of our models that we do not need to be a gigantically high-margin business to be able to afford model training,” OpenAI CEO Sam Altman said on the ‘Invest Like the Best’ podcast.
Last week, OpenAI slashed the price of GPT-5.6 Luna, its fastest and cheapest model for high-volume tasks, by over 80 per cent. Anthropic, on the other hand, has maintained premium pricing for its top-tier Claude models as part of its broader bet that developers will pay extra for safety and precision.
© IE Online Media Services Pvt Ltd
Advertisement
Loading Recommendations...
Advertisement
Live Blog