AI price war intensifies as OpenAI and DeepSeek slash model prices
The latest AI price war compresses frontier model costs in just 48 hours. OpenAI cut its cheapest GPT-5.6 “Luna” pricing by 80% to $0.20/$1.20 per million tokens, and reduced “Terra” by 20% to $2/$12, citing efficiency gains from its own training. A day later, DeepSeek released “V4-Flash” at unchanged rock-bottom rates of $0.14/$0.28 per million tokens—undercutting even OpenAI’s discounted level—while reporting retrained benchmark gains that beat its larger V4-Pro-Preview on agent evaluations.
Anthropic did not cut sticker prices in this window. Instead, it upgraded Opus 4.8 to the more capable Opus 5 at the same $5/$25 rate, emphasizing higher capability per dollar without visible price volatility.
Why it’s happening now: enterprise buyers are getting more cost-sensitive, and model capabilities are converging. When performance gaps shrink, price-per-task and efficiency become the key differentiators—especially as cheaper open-weight models from Chinese labs pressure closed-model vendors.
What traders should watch next: whether other big tech players (e.g., Google or Meta) join with direct pricing moves; whether this accelerates enterprise migration toward multi-model routing; and whether margin pressure shows up in future funding or IPO disclosures. In this AI price war, the market signal is not just discounts, but the push toward “more intelligence per dollar.”
Neutral
This news is about frontier AI model pricing and capability positioning, not a direct crypto protocol change. Still, sustained “AI price war” dynamics can influence crypto sentiment indirectly through enterprise adoption, capex/opex expectations, and the risk appetite of growth/tech investors.
In the short term, aggressive discounts can create a broad “tech cost-cut” narrative that may support risk assets broadly (a sentiment tailwind), but it can also raise uncertainty about margins, which can weigh on markets if investors anticipate funding slowdowns. Historically, similar waves of cloud/model price cuts tend to boost near-term adoption expectations while increasing scrutiny of unit economics.
In the long run, if price-per-task keeps falling and multi-model routing becomes standard, AI infrastructure demand may shift from single-vendor lock-in toward diversified stacks. That can be constructive for crypto-linked “infrastructure and compute” narratives, but the effect is usually gradual and depends on whether enterprise spend shifts into deployable AI workloads rather than just swapping contracts.
Given the article’s lack of direct crypto catalysts (no tokenomics, listings, hacks, or regulatory moves), the expected impact on crypto trading and market stability is best categorized as neutral.