OpenAI API Adds Image Search, GPT-5.4 Bedrock Access, and Granular Billing
OpenAI updates its API in June 2026 with image-enabled web search, moderation scores for responses, and per-minute container billing. Several legacy tools, including reusable prompts and Agent Builder, face upcoming deprecation.
OpenAI introduces several major updates to its developer API in early June 2026, headlined by the addition of image results to the web search tool. Developers using the v1/responses endpoint can now retrieve current, web-grounded visuals alongside standard text results. Additionally, OpenAI releases moderation scores directly within the Responses and Chat Completions APIs, allowing developers to evaluate both input and output safety in a single request without separate API calls.
The company also expands infrastructure and security options by making GPT-5.4 and GPT-5.5 models available through an OpenAI-compatible endpoint in Amazon Bedrock. To improve authentication security, OpenAI launches workload identity federation, which enables trusted systems to exchange external identity tokens for short-lived API access. Furthermore, container session billing shifts to a per-minute model with a five-minute minimum, significantly lowering costs for shorter workloads.
Alongside these new features, OpenAI announces the deprecation of several older platform tools, including reusable prompt objects, the Evals platform, and Agent Builder. The platform also rolls out a redesigned navigation interface and expands its Admin API to handle spend alerts, model allowlists, and data retention settings. For organizations lacking zero-data retention, prompt caching now defaults to a 24-hour retention window instead of an in-memory setting.