OpenAI launches GPT-5 Developer Preview with Modality-Aware API

AI · 6 min read

OpenAI announced the developer preview of GPT-5 today, positioning it as a multimodal foundation model that intelligently routes different input types to specialized submodels. The highlight of the release is the modality-aware API: a single endpoint that analyzes incoming payloads and dispatches them to the most appropriate expert (text, vision, audio, or tabular), returning harmonized outputs.

The preview targets application developers building cross-modal workflows—think voice-driven design tools that can also edit images, or analytics dashboards that accept natural-language queries and images. OpenAI is gating access to the preview with quotas and safety audits; customers must enroll for early access and demonstrate use cases that prioritize transparency and user consent.

OpenAI emphasized safety and interpretability: the API returns a brief routing log with each response that shows which experts were used and why, plus confidence metrics. Pricing and general availability timelines were not finalized, but the company said wider rollout will follow iterative feedback from the developer preview program.