OpenAI plans major price cuts and vision API for devs

If you have tried building on top of OpenAI’s models, you already know the real killer isn't model latency—it's the token bill.

Because the current API is completely stateless, developers have to resend the entire conversation history with every single prompt. A single one-page document analysis on GPT-4 can easily cost 10 cents, and those micro-transactions add up to bankrupting cloud bills fast.

According to Reuters, OpenAI is finally addressing this bottleneck. At its first-ever developer conference in San Francisco on November 6, the company plans to unveil massive updates designed to make building applications significantly cheaper and faster.

Stateful APIs and built-in vision

The marquee update is the addition of memory storage for developer tools, effectively creating a stateful API.

By retaining conversation logs on OpenAI’s side instead of forcing apps to transmit duplicate context on every call, sources say this change could theoretically slash developer costs by up to 20 times.

Alongside memory, OpenAI is rolling out its Vision API to developers. Weeks after launching visual capabilities for consumer ChatGPT, third-party developers will finally be able to feed images directly into their apps, targeting use cases from medical screening to entertainment.

It is all part of CEO Sam Altman's grand strategy to turn a viral consumer chatbot into an indispensable enterprise developer platform.

Moving past the plugin slump

OpenAI needs this developer love badly right now.

The company is chasing aggressive revenue targets, with executives previously eyeing $200 million this year and $1 billion by 2024. While consumer adoption skyrocketed, its B2B developer plays have hit unexpected friction.

Earlier this year, OpenAI rushed out ChatGPT plugins, hoping to recreate Apple's App Store playbook against rivals like Google’s Bard. But after an initial wave of hype, user retention fell off a cliff. Even Scholar AI, one of the top 30 plugins, recently saw just around 7,000 daily active users.

At the same time, founders are getting nervous. While PitchBook reports over $20 billion poured into AI startups this year, VCs are warning portfolio companies against building entirely on OpenAI. Many developers have started hedging their bets by diversifying into Meta's open-source Llama and rival foundation models.

Why it matters

For developers and founders, a 20-fold theoretical cost reduction completely rewrites the unit economics of AI software. Long-running conversational agents, customer service tools, and continuous document processing that were previously cost-prohibitive might suddenly become viable, profitable businesses.

For the broader industry, OpenAI is making a ruthless counter-offensive. Meta’s open-source weights have been winning hearts by offering cost control and flexibility. By bringing down its own proprietary API toll and wrapping vision into the stack, OpenAI is putting intense margin pressure on both open-source alternatives and cloud giants like Google.

We think shifting focus away from hollow storefronts like plugins and toward developer infrastructure is the smartest move OpenAI has made all year. However, the core hazard remains intact. As OpenAI bundles memory and multimodal tooling natively into its API, it risks steamrolling the very middleware startups it is courting. Winning long-term developer loyalty requires Altman to prove OpenAI wants to be an engine, not the entire vehicle.