This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(105.60000000000001, 55.00000000000001%, 20%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EA%3C/text%3E%3C/svg%3E)
OpenAI released their https://developers.openai.com/api/docs/guides/websocket-mode/, for faster AI responses.
An Azure service that provides a hybrid, multi-cloud management platform for APIs.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(316.8, 39%, 40%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ESM%3C/text%3E%3C/svg%3E)
Hi andrew
OpenAI added WebSocket transport support to the Responses API, allowing clients to maintain a persistent WebSocket connection (wss://api.openai.com/v1/responses) instead of issuing repeated stateless HTTP requests. This is documented in the official OpenAI API guide under WebSocket Mode. [developers...openai.com]
WebSocket mode is designed to address performance overhead caused by repeatedly opening HTTP connections and resending full conversation context. With WebSocket mode, the connection remains open and each subsequent turn sends only incremental input along with a previous_response_id. This significantly reduces continuation overhead in long‑running workflows. [developers...openai.com]
According to OpenAI, for workflows involving 20 or more sequential tool calls, WebSocket mode can deliver up to approximately 40% faster end‑to‑end execution compared to traditional HTTP‑based calls. This improvement is attributed to reduced connection setup time and minimized payload size per turn. These figures are explicitly stated in OpenAI documentation and corroborated by independent reports.
OpenAI positions WebSocket mode primarily for agentic and tool‑heavy workflows, such as coding agents, automation pipelines, orchestration loops, and multi‑step reasoning systems. It is not marketed as a way to speed up single‑turn chat responses, but rather to optimize long chains of model‑tool interactions.
In WebSocket mode, each interaction is initiated by sending a response.create event over the open connection. Subsequent turns reference the prior response using previous_response_id and include only new user messages or tool outputs.
OpenAI also documents an optional “warm‑up” call (generate: false) to prepare server‑side state and further reduce latency for the next generated turn. [developers...openai.com]
WebSocket mode is explicitly documented as being compatible with Zero Data Retention (ZDR) and supports store=false, making it suitable for privacy‑sensitive and regulated enterprise workloads. OpenAI confirms that persistent connections do not require response storage to be enabled
Known limitations:
OpenAI‑documented and publicly reported constraints include a maximum WebSocket connection lifetime of 60 minutes and single‑request processing per connection. Applications requiring parallel execution must establish multiple WebSocket connections. These limitations are discussed in third‑party summaries based on OpenAI’s behavior and documentation
OpenAI offers two WebSocket‑based approaches: Responses API WebSocket mode, which targets tool‑heavy agent workflows, and the Realtime API, which is optimized for low‑latency audio and voice interactions. WebSocket mode applies specifically to the Responses API and should not be confused with the Realtime audio/voice endpoint
References:
I Hope this helps. Do let me know if you have any further queries.
Thankyou!
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(176, 16%, 27%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ESC%3C/text%3E%3C/svg%3E)
Hi andrew,
Did you get any chance to review the above response. Do let me know if you have any further queries.
Thank you!
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(204.8, 15%, 29%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EMM%3C/text%3E%3C/svg%3E)
Hi andrew
There is no official announcement for enabling WebSocket on response Apis yet.
All latest updates are recorded here in What is new page.
For now, it is only available on Audio Realtime models as my colleague pointed out.
Feel free to bring supportability on response api as feature request in Azure Feedback forum.
Thank you for using Azure Services.
Hi andrew
Please take a minute to accept this answer as verified one if you found my answer useful.
Thank you.
Yes, WebSocket mode is supported on Azure for OpenAI models. The Realtime API, which operates via WebSockets, allows for fully asynchronous streaming communication between the end user and the model. This API can be accessed through a secure WebSocket connection to the /realtime endpoint of your Azure OpenAI resource. It is designed to facilitate low-latency interactions, making it suitable for applications that require real-time responses.
To use WebSocket mode, you need to construct a request URI that includes your Azure OpenAI resource endpoint and the appropriate API version and model deployment parameters. Authentication can be done using Microsoft Entra or an API key.
References: