OpenAI Developer Day Highlights: Custom GPTs, GPT‑4 Turbo, and New Assistant API

OpenAI announced the launch of custom GPTs at its first Developer Day, enabling users to create personalized assistants without any programming by simply describing desired capabilities.

The new GPT‑4 Turbo model expands the context window to 128 k tokens, updates knowledge to April 2023, adds multimodal vision, DALL·E 3, text‑to‑speech, and offers significantly reduced API pricing.

A GPT Store will allow creators to share and monetize their custom GPTs, while the new Assistant API introduces persistent, unlimited‑length threads, retrieval of external knowledge, a code interpreter, and multi‑function calling.

“We see a new abstraction layer in computing,” said OpenAI founding member Andrej Karpathy.

Demonstrations showed how anyone can build specialized GPTs—such as a startup mentor—through a conversational builder, with options to enable internet access, image generation, and code execution, and to share the GPT internally or publicly.

Pricing updates cut GPT‑4 Turbo input costs to one‑third and output costs to half of previous rates; the Assistant API’s code interpreter is billed per use, with a free trial period available.

“We focus on the best infrastructure, you focus on the best models,” said Microsoft CEO Satya Nadella.

These announcements position AI as a versatile tool for developers across domains, marking a significant evolution in how AI models are customized and deployed.