That puts Deepseek’s flagship model far below the pricing of competing frontier systems. GPT-5.5 charges $5 per million input tokens and $30 per million output tokens, while Anthropic’s Opus 4.7 costs $5 for input and $25 for output. Deepseek’s lower-tier V4 Flash model is cheaper still.
The pricing gap becomes especially significant for agentic AI systems and large-scale workflows that consume massive numbers of output tokens. Both Deepseek models also support context windows up to one million tokens and can generate as many as 384,000 output tokens. The company additionally supports OpenAI and Anthropic API formats, reducing friction for developers looking to switch providers.
While raw token pricing offers one comparison point, the report notes that actual usage costs depend heavily on how efficiently models consume tokens during real-world tasks. Some competing systems advertise lower rates while using substantially more tokens per interaction, which can offset headline savings.
Deepseek’s models still trail top-tier systems like GPT-5.5 and Opus 4.7 in overall frontier performance, according to the report, but the scale of the pricing difference could reshape purchasing decisions for businesses under pressure to manage AI infrastructure costs. As companies move beyond experimentation and begin deploying AI more broadly, many are becoming increasingly focused on whether a cheaper model is “good enough” rather than whether it leads every benchmark.
The move also highlights a growing divide in how AI companies are approaching the market. OpenAI and Anthropic continue operating under enormous revenue pressure as both companies move toward potential IPOs and spend heavily on training and infrastructure. Deepseek, which is entering its first funding round, appears to be pursuing a different strategy centered on aggressive pricing and developer adoption.
The permanent discount adds new pressure to an AI market already facing rapid commoditization around inference pricing. As model capabilities become more competitive across providers, pricing, compatibility, and deployment costs are starting to matter more for developers deciding which platforms to build on long term.
This analysis is based on reporting from The Decoder.
Image courtesy of Cath Virginia / The Verge.
This article was generated with AI assistance and reviewed for accuracy and quality.
