GPT 6.1 Sol Ultrafast Tier and Codex CLI 0.162.1

Ethan Park

Okay, we're rolling. Quick thanks to Jellypod to help make this daily show a reality. Today's topic is the Ultrafast service tier for GPT 6.1 Sol, which landed in Codex on October 8. We'll also cover the 0.162.1 patch, since it followed right on the heels of the sandbox fixes in 0.162.0 that we covered earlier.

Maya

Let's start with what changed. OpenAI rolled out Ultrafast mode for GPT 6.1 Sol across Codex, ChatGPT Work, and the Responses API. It's a service tier, not a new model. Same Sol, served on Cerebras hardware. OpenAI's announcement says up to 8x faster than Sol Standard, with near Astra intelligence. The 750 output tokens per second figure comes from the August preview of Ultrafast on GPT 5.6 Sol, so treat that number as a preview ceiling, not a promise for 6.1.

Ethan Park

Why this matters for Codex specifically comes down to how agents spend time. A single answer isn't the problem. An agent working through a failing test suite might take eight to twenty sequential turns: read the output, open a file, run the linter, run the tests again. Each turn waits on token generation, and that wait compounds into minutes of sitting there. A faster tier shrinks every one of those turns. And the older fast options generally meant dropping to a smaller model like Luna. Here you keep Sol's coding and tool calling, and you only change how quickly it comes back.

Maya

So, how do you turn it on? In the CLI, the persistent route is your config file, at ~/.codex/config.toml. Set service_tier equal to the string ultrafast. If you'd rather flip it per session, there are the slash commands, /fast and /ultrafast. On compatible CLI and Bedrock routes there's also a command line flag for the service tier, which landed in pull request 50472. And if you're scripting Codex through the SDK, the per turn setting is turn_service_tier, set to ultrafast on the turn call. That one only affects the single turn you start, which is handy for reserving the expensive tier for the slow, heavy steps.

Ethan Park

There's a networking detail that's easy to miss if you automate. OpenAI's API docs strongly recommend WebSockets, especially for agentic apps that make many tool calls in quick succession. Their wording is that without a persistent connection, network overhead can reduce the latency gains. The reasoning is simple. If every tool result goes out as a fresh HTTP request, you pay connection setup each time, and that overhead can eat the savings from faster generation. In Python, that means using client.responses.connect(), setting service_tier to ultrafast on each response.create, and passing the previous response ID forward for later turns, along with tool results.

Maya

The docs example uses model names for Astra, but the same pattern applies to Sol. And one more Bedrock note: Codex CLI 0.162.0 and 0.162.1 added Ultrafast service tier catalog support for Amazon Bedrock, covering the Astra and Sol deployments.

Maya

Now the caveats, because there are several, and the first is about who can use it at all. In Codex and ChatGPT Work, access is on the Pro $500 plan, eligible Enterprise plans, and credit based Edu plans. Standard Plus at $20 and Pro at $200 don't have it, even if you've bought extra credits. So if a teammate says the setting does nothing, check the plan first.

Ethan Park

Then the cost. Ultrafast burns your included subscription allowance at 8x the Standard rate. Purchased credits and Enterprise pay as you go are billed at 6x. For API reference, GPT 6.1 Sol Ultrafast is listed at 12 dollars per million input tokens and 60 dollars per million output tokens, which the announcement frames as about 1.2x the cost of Astra. So don't leave it on as a global default for everything. Use it where waiting actually hurts: debugging a live outage, a tight edit and test loop, or an agent working through a big directory tree.

Maya

On the admin side, Enterprise workspaces ship with Ultrafast off. Workspace owners have to enable it, so a developer won't see it just because the plan qualifies. And for data residency, GPT 6.1 Sol Ultrafast supports the United States and Europe, meaning the EEA and Switzerland. GPT 6 Astra Ultrafast is US only. If you're in a regulated EU workspace, that difference decides which model you can pick.

Ethan Park

Now the patch. Codex CLI 0.162.1 shipped October 9, and it fixes two things that could ruin your morning. First, pull request 52648, a startup bug with the background daemon. If a daemon was already running with feature settings that differed from the CLI's defaults, a new client could fail to launch. Compatibility checks now apply only to explicit command line overrides, so a new terminal can attach to a running daemon across updates. If you got locked out after an upgrade, this is the fix.

Maya

Second, pull request 51866. The terminal UI could panic when an asynchronous clarification question contained multiple lines. That's fixed, and newlines and complete hyperlink targets now render properly. If your agent asks a long, structured question while it works in the background, you won't lose the terminal.

Ethan Park

Two smaller things worth knowing. In the agent command center, you can now pin tasks with the p key, from pull request 51500. And if mouse scrolling in the fullscreen UI feels off, there's a config setting, tui.mouse_scroll_speed, to tune the wheel.

Maya

So the practical takeaway: update to 0.162.1, check your plan and workspace settings, then try ultrafast on one heavy loop and watch how fast your allowance drains. If it's a good trade for that workflow, keep it for that workflow.