GPT-5 pricing sparks market shift

OpenAI's GPT-5 arrives with aggressive pricing that could reshape the AI API market.

GPT-5 pricing triggers a new AI price war

OpenAI released GPT-5, its latest flagship model, just days after offering two open source models. The API pricing is $1.25 per 1 million input tokens and $10 per 1 million output tokens, plus a cache fee of $0.125 per 1 million tokens. The structure mirrors Google Gemini 2.5 Pro basic pricing, though Google charges more when use climbs past a high threshold. On a per-token basis, GPT-5 undercuts several rivals on typical coding tasks.

OpenAI faces competition from Claude Opus 4.1, which starts at $15 per 1 million input tokens and $75 per 1 million output tokens, though discounts exist for prompt caching and batch processing. Early testers describe GPT-5 as strong across many tasks, especially coding, while reviewers note that benchmark results are mixed across different tests. The company has big infrastructure commitments, including a $30 billion-per-year Oracle contract, and rivals like Meta and Alphabet are earmarking tens of billions for AI capacity. If rivals respond with lower prices, the sector could drift toward a price war that reshapes incentives for developers and tool makers.

Pricing is becoming a core feature of AI products, not just a support cost. When prices fall, developers can experiment more and build broader use cases, which in turn drives broader adoption. That dynamic could reward efficiency and caching strategies that lower effective costs for customers. At the same time, a sustained price war would compress margins for large providers and could slow the pace of innovation if profits shrink.

Investors will watch pricing moves as a signal of how the underlying economics of AI access are evolving. If the market settles into lower prices, startups and smaller firms may flourish; if prices rebound, the market could consolidate around a few dominant players. The scale of current investments in data centers and capacity means any long-term price drop has to be balanced against the need to fund ongoing research and infrastructure.

Highlights

The market will test which cost models actually reward long-term value rather than short-term wins.