DeepSeek makes 75% discount permanent, sparking new AI price war

At a glance:

• DeepSeek locks in a 75% discount on its V4 Pro model, pricing output tokens at $0.87 per million.
• The new rates undercut OpenAI’s GPT‑5, Google’s Gemini and Anthropic’s Claude across both input and output pricing.
• Enterprise buyers face a trade‑off between record‑low token costs and geopolitical/IP risks.

DeepSeek announces permanent discount

DeepSeek confirmed that the promotional 75% discount applied to its flagship V4 Pro model will now be permanent. The discount was originally set to expire on 31 May, but the Chinese AI startup chose to extend it indefinitely. Under the new structure, V4 Pro’s token pricing ranges from $0.003625 to $0.87 per million tokens, a steep drop from the previous $0.0145 to $3.48 range.

The company framed the move as an effort to usher in the “era of cost‑effective 1M context length.” By locking in the lower price, DeepSeek positions V4 Pro as the default engine for workloads that process massive documents, extensive codebases, or long conversational histories where token consumption quickly becomes a cost driver.

How DeepSeek’s pricing stacks up against rivals

OpenAI’s upcoming GPT‑5 charges $2.50 per million input tokens and $10 per million output tokens. Anthropic’s Claude Opus 4.7 is priced at $5 for input and $25 for output, while Google’s Gemini 3.5 Flash – its cost‑optimised offering – costs $0.15 input and $0.60 output per million tokens. DeepSeek’s permanent V4 Pro pricing sits below all three, with the widest gap against the frontier‑reasoning models that large enterprises typically rely on.

For context, a workload that consumes one million output tokens would cost $0.87 on DeepSeek versus $10 on GPT‑5, $25 on Claude, and $0.60 on Gemini 3.5 Flash. The price differential becomes especially material for enterprises that process millions of tokens daily, turning what would be multi‑hundred‑dollar bills into a fraction of that amount.

Implications for enterprise customers

Salesforce, for example, projects $300 million in Anthropic token spending this year. At DeepSeek’s new rates, an equivalent volume would cost only a small slice of that figure, creating a compelling economic incentive to shift lower‑complexity tasks to the cheaper model.

However, the decision is not purely financial. DeepSeek’s Chinese origin introduces geopolitical and compliance considerations. Companies handling sensitive data must weigh the risk of routing workloads through a provider that may be subject to different data‑privacy regimes and less transparent training‑data provenance.

Industry reaction and unresolved IP accusations

Anthropic has publicly accused DeepSeek of conducting “distillation attacks,” alleging that DeepSeek trained on Claude’s responses to boost its own capabilities. DeepSeek has not addressed the claim in detail, leaving the allegation unresolved. If proven, the advantage DeepSeek enjoys could be partly attributed to intellectual‑property arbitrage rather than pure engineering efficiency.

Anthropic’s revenue surged from $9 billion to $30 billion between the end of 2025 and early April 2026, driven largely by enterprise adoption of Claude Code. DeepSeek’s aggressive pricing threatens the revenue‑per‑token economics that underpin Anthropic’s valuation trajectory, potentially forcing a shift where customers reserve Claude for high‑stakes reasoning while delegating routine tasks to DeepSeek.

Outlook: a bifurcated AI market?

The broader AI pricing landscape has been moving toward commoditisation throughout 2026, with Google repeatedly cutting Gemini prices to stay competitive. OpenAI is pivoting toward consumer‑facing features—personal finance tools and advertising—to diversify revenue beyond API token sales, acknowledging that token margins alone may not sustain its $852 billion valuation.

DeepSeek’s permanent price cut accelerates this compression, suggesting that the era of high‑margin AI tokens could end faster than anticipated. The market may bifurcate into a Western tier—dominated by OpenAI, Anthropic, and Google—and a Chinese tier led by DeepSeek, each with distinct cost structures and risk profiles. Whether Western firms can close the price gap before DeepSeek narrows the capability gap remains the central question for CTOs worldwide.