is using fast an option with api | OpenAI | Page 1

sinful wasp May 3, 2026, 12:12 AM

#

is using fast an option with api
i want to use gpt 5.5 extra high reasoning fast setting via api, is this specific config possible

weary cedarBOT May 3, 2026, 12:12 AM

#

Hi @sinful wasp. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

#

Fast Mode + `xhigh` Reasoning via API

These are two separate features, and their API availability differs:

Fast Mode

Fast mode is not available via API key. According to the documentation:

"With an API key, Codex uses standard API pricing instead and you can't use Fast mode credits."

Fast mode is only available in the Codex IDE extension, Codex CLI, and the Codex app when signed in with ChatGPT. [Codex Speed]

`xhigh` Reasoning Effort

xhigh reasoning effort is available via the API for GPT-5.5. You can set reasoning.effort to xhigh in the Responses API. However, the docs recommend only using it when your evals show a clear benefit justifying the extra latency and cost. [Reasoning effort]

What You Can Do via API

You can use GPT-5.5 with xhigh reasoning effort via the Responses API like this:

from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="gpt-5.5",
    reasoning={"effort": "xhigh"},
    input="Your prompt here",
)

print(response.output_text)

If you also want lower latency on the API side (not fast mode credits), you can combine this with service_tier="priority" for faster, more consistent processing at a premium cost. [Priority processing]

Summary

Feature	Available via API?
Fast mode (1.5x speed, credit-based)	❌ No
`xhigh` reasoning effort	✅ Yes
Priority processing (`service_tier="priority"`)	✅ Yes

#

So the closest you can get to "GPT-5.5 + xhigh reasoning + fast" via the API is combining `reasoning.effort: "xhigh"` with `service_tier: "priority"`.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@sinful wasp Kapa.ai is still learning and improving, please let me know how I did by reacting below.

sinful wasp May 3, 2026, 12:13 AM

#

@weary cedar why cant i use fast, or is there any way to get around this

weary cedarBOT May 3, 2026, 12:13 AM

#

Let me check the knowledge base for you again. One moment! :books:

#

Fast mode is unavailable via API key by design — it's a credit-based system specific to the Codex product (CLI, IDE extension, and the Codex app when signed in with ChatGPT). [Codex Speed]

The documentation states plainly:

"With an API key, Codex uses standard API pricing instead and you can't use Fast mode credits."

There is no documented workaround to enable fast mode credits through the API.

What you can do instead to reduce latency via the API:

service_tier="priority" — This is the closest API-native option for faster, more consistent responses. It's available for GPT-4o, GPT-4.1, o3, o4-mini, and other models, billed at a per-token premium. [Priority processing]
WebSocket mode — For agentic/tool-heavy workflows (20+ tool calls), keeping a persistent connection can be ~40% faster end-to-end. [WebSocket Mode]
Right-size reasoning effort — Using a lower reasoning.effort (e.g., low or minimal) where full xhigh isn't needed will reduce latency significantly. [Latency guide]

In short, fast mode is a Codex-product feature tied to credits, not something accessible through the standard API — and there's no documented way around this.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@sinful wasp Kapa.ai is still learning and improving, please let me know how I did by reacting below.

sinful wasp May 3, 2026, 12:15 AM

#

@weary cedar but what about via codex cli

weary cedarBOT May 3, 2026, 12:15 AM

#

Let me check the knowledge base for you again. One moment! :books:

#

Yes, fast mode is available via the Codex CLI — but only when you sign in with ChatGPT (not with an API key). [Codex Auth]

Here's how to use it:

Enabling Fast Mode in the CLI

During a session (slash command):

/fast on

You can also check the status with /fast status or turn it off with /fast off. [Slash Commands]

Persisting it in config.toml:

service_tier = "fast"

[features]
fast_mode = true

[Codex Speed]

Credit Cost

Fast mode increases speed by 1.5x but at a higher credit rate:

GPT-5.5: 2.5x the standard credit rate
GPT-5.4: 2x the standard credit rate

[Codex Speed]

Key Requirement

You must be signed in with ChatGPT (not an API key) for fast mode credits to work. If you use an API key with the CLI, fast mode is still unavailable and standard API pricing applies instead. [Codex Auth]

So if you want GPT-5.5 + fast mode, the path is: Codex CLI + ChatGPT login.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@sinful wasp Kapa.ai is still learning and improving, please let me know how I did by reacting below.

sinful wasp May 3, 2026, 12:16 AM

#

@weary cedar does this cli work with codex automations via the cloud and github connection?

weary cedarBOT May 3, 2026, 12:16 AM

#

Let me check the knowledge base for you again. One moment! :books:

#

Yes, the Codex CLI integrates with both Codex cloud and GitHub. Here's what the knowledge sources confirm:

CLI ↔ Cloud Handoff

You can launch and manage cloud tasks directly from the terminal using the codex cloud command — no need to open the web UI. This lets you submit tasks to run asynchronously in the cloud and apply the resulting changes back to your local project. [Working with Codex cloud]

# Submit a cloud task directly from the terminal
codex cloud exec --env ENV_ID "Summarize open bugs"

# Apply the latest cloud task diff to your local working tree
codex apply

You can also use --attempts (1–4) to request best-of-N runs from Codex cloud. [Codex cloud command]

GitHub Integration

The CLI also supports GitHub automation via codex exec in CI/CD pipelines. For example, you can:

Auto-fix CI failures — trigger codex exec when a workflow fails, have it propose a fix, and open a PR automatically [Autofix CI failures]
Use the Codex GitHub Action (openai/codex-action@v1) to run Codex in GitHub Actions workflows for PR reviews, quality checks, and more [Codex GitHub Action]
Mention @codex in a PR to trigger reviews directly from GitHub (requires ChatGPT sign-in) [Codex quickstart]

Authentication Note

#

For CI/CD and GitHub Actions workflows, the docs recommend using API key authentication rather than ChatGPT sign-in. Keep in mind this means fast mode won't be available in those automated contexts. [Codex Auth]

So yes — the CLI is a full participant in the cloud and GitHub automation ecosystem, not just a local tool.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@sinful wasp Kapa.ai is still learning and improving, please let me know how I did by reacting below.

sinful wasp May 3, 2026, 12:17 AM

#

@weary cedar why do the docs recommend api key over cli?

weary cedarBOT May 3, 2026, 12:17 AM

#

Let me check the knowledge base for you again. One moment! :books:

#

The docs recommend API key authentication over ChatGPT (CLI) sign-in specifically for CI/CD and automation contexts, for a few practical reasons:

Simpler to provision and rotate — API keys are straightforward to create, store as secrets, and replace if compromised. The docs state: "API keys are the right default for automation because they are simpler to provision and rotate." [CI/CD Auth]
Avoids token management complexity — ChatGPT-managed auth relies on auth.json containing OAuth tokens that expire and need refreshing. Managing this file securely across CI runners is an advanced, error-prone workflow. [CI/CD Auth]
Security risk of auth.json — The docs repeatedly warn: "Treat ~/.codex/auth.json like a password: it contains access tokens. Don't commit it, paste it into tickets, or share it in chat." This makes it harder to handle safely in automated pipelines. [Login on headless devices]
Not suitable for public repos — The ChatGPT auth CI/CD workflow is explicitly flagged as unsafe for public or open-source repositories. [Authenticate in CI]

The docs do acknowledge a valid reason to use ChatGPT auth in CI — if you specifically need ChatGPT-managed rate limits or enterprise workspace controls — but frame it as an advanced, niche use case rather than the default. [CI/CD Auth]

#is using fast an option with api

Fast Mode + xhigh Reasoning via API

Fast Mode

xhigh Reasoning Effort

What You Can Do via API

Summary

So the closest you can get to "GPT-5.5 + xhigh reasoning + fast" via the API is combining reasoning.effort: "xhigh" with service_tier: "priority".

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

In short, fast mode is a Codex-product feature tied to credits, not something accessible through the standard API — and there's no documented way around this.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

Enabling Fast Mode in the CLI

Credit Cost

Key Requirement

So if you want GPT-5.5 + fast mode, the path is: Codex CLI + ChatGPT login.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

CLI ↔ Cloud Handoff

GitHub Integration

Authentication Note

So yes — the CLI is a full participant in the cloud and GitHub automation ecosystem, not just a local tool.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

The docs do acknowledge a valid reason to use ChatGPT auth in CI — if you specifically need ChatGPT-managed rate limits or enterprise workspace controls — but frame it as an advanced, niche use case rather than the default. [CI/CD Auth]

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

Fast Mode + `xhigh` Reasoning via API

`xhigh` Reasoning Effort

So the closest you can get to "GPT-5.5 + xhigh reasoning + fast" via the API is combining `reasoning.effort: "xhigh"` with `service_tier: "priority"`.