OpenAI doubles GPT-5.5 prices while DeepSeek undercuts by 90% with open-source models, forcing developers to choose between premium APIs and cheap infrastructure

On April 23–24, OpenAI raised GPT-5.5 pricing to $5 per million input tokens and $30 per million output tokens (double GPT-5.4's rates), while DeepSeek released V4-Pro at $1.74 input and $3.48 output, and V4-Flash at $0.14 input and $0.28 output—all with open-source weights under MIT license on Hugging Face.

OpenAI is selling an integrated product stack (model + agent tools + computer-use features + single vendor support) as a premium offering; DeepSeek is betting frontier AI becomes commodity infrastructure like Linux, using efficient Mixture-of-Experts architecture (49 billion active parameters in V4-Pro) that runs on cheaper hardware and can be self-hosted or embedded in products.

For developers building coding agents and scaling inference pipelines, the "comfortable middle" tier is disappearing—they now choose between paying for OpenAI's integrated enterprise product or running DeepSeek's open models themselves, with the price gap large enough that many production systems will route different tasks to each option.

DeepSeek V4 is text-only at launch (image and video support in progress), so for workloads requiring multimodal reasoning, it is not yet a direct replacement for GPT-5.5; V4-Flash carries a launch discount through May 5, 2026, and both models can run on mid-size GPU clusters that smaller teams can afford.