OpenAI used its first developer conference to unveil a broad expansion of its platform, led by GPT-4 Turbo with a 128,000-token context window, lower API prices, new developer tools and a system for users to build customized versions of ChatGPT without writing code. The company's DevDay announcement on Monday marks its most aggressive effort yet to turn the technology behind ChatGPT into a general-purpose software platform.

GPT-4 Turbo can process far more material in a single prompt than the standard GPT-4 models available earlier this year, while OpenAI says the model is also more capable at following instructions and working with structured outputs. The company priced GPT-4 Turbo input tokens at one-third the cost of GPT-4 and output tokens at half the prior price, a change aimed directly at developers trying to build commercially viable applications around large language models.

A larger context window changes what developers can attempt

The 128K context window is roughly equivalent to hundreds of pages of text, allowing developers to place much larger documents, code bases or conversations into a model request. That does not guarantee perfect recall across every token, but it changes the practical ceiling for document analysis, retrieval workflows and long-running applications.

OpenAI is also adding JSON mode, improved function calling, reproducible outputs through a seed parameter and access to log probabilities. Those features are less visible than a new model name but matter to software teams because they make model behavior easier to integrate into ordinary application logic. The company's existing GPT-4 release in March established the multimodal and reasoning baseline; GPT-4 Turbo is designed to make that capability cheaper and easier to use at scale.

The company also introduced an Assistants API with built-in retrieval, code execution and function-calling tools. Instead of requiring developers to assemble every component of a conversational application themselves, OpenAI is offering more of the orchestration layer directly.

Custom GPTs push ChatGPT toward an application marketplace

For consumers and organizations, the more visible announcement is GPTs: custom versions of ChatGPT configured with instructions, uploaded knowledge and selected capabilities. In a separate product announcement, OpenAI said Plus and Enterprise users can create GPTs for specific tasks and share them with others. The company plans to launch a GPT Store later this month where public GPTs can be discovered.

The idea lowers the barrier between using an AI assistant and building a small application. A teacher might configure a tutoring GPT around a curriculum, a company might create an internal policy assistant, or a specialist might combine instructions with proprietary reference material. OpenAI's ChatGPT release notes describe the November 6 rollout as a new way to tailor ChatGPT for work, home and specialized tasks.

The strategy also extends the enterprise direction OpenAI announced in August. ChatGPT Enterprise added administrative controls, stronger privacy commitments, longer context and unlimited higher-speed GPT-4 access for organizations. GPTs now offer another layer of customization that can remain internal to a company or be shared publicly.

Multimodal tools move into the same developer stack

DevDay also brings GPT-4 Turbo with vision, DALL·E 3 image generation and text-to-speech into the API. That makes it possible for one application to combine text reasoning, image understanding, image generation and spoken output using a common vendor and account infrastructure.

OpenAI has been moving toward that convergence for months. Its DALL·E 3 announcement described an image model designed to follow detailed language instructions more faithfully and to work closely with ChatGPT as a prompt-refinement interface. Opening DALL·E 3 and vision capabilities more broadly to developers turns those consumer-facing features into building blocks for outside products.

The expanded platform is also accompanied by Copyright Shield, under which OpenAI says it will defend customers and pay costs incurred in certain copyright-infringement claims involving generally available OpenAI systems. The commitment applies to ChatGPT Enterprise and the API, though not to every use or modification.

Price becomes a competitive weapon

The most immediate commercial effect may come from pricing. Large language model applications can become expensive when they process large prompts, serve many users or repeatedly call models within one workflow. Cutting token prices reduces the cost of experimentation and makes some previously marginal applications more realistic.

TechCrunch's contemporaneous DevDay report highlighted both the model upgrades and the pricing changes, along with an experimental program for GPT-4 fine-tuning. OpenAI is also increasing rate limits for established GPT-4 API customers, another signal that the company is preparing for higher-volume production use.

The announcements place pressure on rival model providers to compete not only on benchmark performance but on context length, developer tooling, reliability, data controls and cost. At the same time, OpenAI is taking on more responsibility for the application layer itself. As it supplies retrieval, code execution, custom assistants and a GPT marketplace, it moves closer to areas where many startups have been building products on top of its models.

From model provider to platform company

DevDay's larger message is that OpenAI no longer wants to be evaluated only by the intelligence of its latest model. It is assembling the components of a platform: models, tools, storage and retrieval, multimodal APIs, enterprise controls, customization, a distribution channel and lower unit costs.

That shift could make OpenAI more deeply embedded in software development, but it also increases execution demands. Developers will judge the platform on uptime, version stability, predictable pricing, privacy and whether new abstractions remain useful as models change.

For now, GPT-4 Turbo's 128K context window and cheaper pricing are substantial technical changes, while GPTs represent a bet that millions of users can become lightweight application builders. Together they make the company's first DevDay less a product refresh than a declaration that generative AI is becoming an application platform in its own right.