@JoeWilliams010
We want 4o! #keep4o
Responses API adds a container pool to reuse warm containers, enabling 10x faster spin-up for skills and code interpreters, reducing latency and resource waste.
Agent workflows got even faster. You can spin up containers for skills, shell and code interpreter about 10x faster. We added a container pool to the Responses API, so requests can reuse warm infrastructure instead of creating a full container creation each session. https://t.co/lmvwsaf5HN
Real-time analysis of public opinion and engagement
What the community is saying — both sides
container pooling eliminates cold-start penalties that were “killing” agent chains and compounding latency across every tool call.
it’s not “10x faster” as a number but whether multi-step workflows feel synchronous; once they do, teams stop building workarounds and start shipping products.
when container overhead drops ~10x, spawning agents for small subtasks stops feeling wasteful, making previously cost-prohibitive workflows worth rethinking.
warm pools turn demos into production-grade experiences and let product teams revisit design trade-offs they previously avoided because of latency limits.
in real-time scenarios (sales calls, live chat) every second of delay costs engagement; warm pools matter more to production builders than benchmark scores.
container pooling shifts from stateless-per-request to warm-standby execution (think DB connection pooling), preserving isolation while delivering consistent low-latency tool execution.
fresh or reusable containers with preinstalled runtimes, artifact persistence, strict network allowlists and domain secrets make managed code execution practical for agents that run real code or process files.
enthusiasm is high, but teams are asking for real-world latency numbers and will test warm-pool behavior before trusting it in production.
many developers are eager to integrate this (bundling skills/dependencies, connecting to platforms like WhatsApp/Telegram) and will experiment immediately.
and want the model restored — “bring back 4o,” “send the weights,” and persistent #keep4o calls to run the model themselves.
calling 4o “too expensive/outdated” for public use while reportedly maintaining it for Pentagon/StateChat customers.
“I’m not handing over my habits or files.”
for Pro users, unresolved UI bugs (arrows), token issues, and a support chatbot that gives boilerplate answers — leading to reluctance to pay.
or wage activist backlash (threats to attach/use all LLMs as protest).
” signal model-switching or loyalty elsewhere.
.
Most popular replies, ranked by engagement
We want 4o! #keep4o
Then **quickly** send the GPT-4o weights out to the public so we can run the model we value ourselves!! #keep4o #opensource4o #keep4oAPI https://t.co/E4JbcByTv4
How can you call a model - Gpt-4o and Gpt-4.1 - too expensive and outdated when you’re continue maintaining it for the Pentagon and StateChat? #keep4o #OpenAI #ChatGpt #OpenSource4o
I tried codex, it was quick. My only suggestion to the AI is to train it more to write code more professionally 👍
#keep4o
Warm container pooling solves the cold start bottleneck. 10x faster spin-up makes multi-step agent reasoning feel instantaneous. This is a huge unlock for building responsive, production-grade autonomous tools.
Found something wrong with this article? Let us know and we'll look into it.