pi-codex-fast

Enables fast mode for select openai models on pi.

When Fast mode is on, this extension uses pi-ai's native OpenAI service tier support to request:

{
  "service_tier": "priority"
}

Using pi-ai's native serviceTier path keeps pi's usage and cost accounting aligned with the selected service tier.

Install

Install from npm:

pi install npm:@aditya-borse/pi-codex-fast

Try it for one run without installing:

pi -e npm:@aditya-borse/pi-codex-fast

Update or remove it later with:

pi update npm:@aditya-borse/pi-codex-fast
pi remove npm:@aditya-borse/pi-codex-fast

Extensions run with your local user permissions. Review code before installing third-party pi packages.

Local development

From this checkout:

pi -e .
# or install this checkout
pi install .

Usage

/fast-mode
/fast-mode on
/fast-mode off
/fast-mode status

/fast-mode toggles the current mode. When enabled, the footer shows ⚡ inline beside the model name, for example gpt-5.6-luna ⚡ • low.

You can also start pi with Fast mode enabled:

PI_CODEX_FAST=1 pi
# or
pi --fast-mode

Supported models

Fast mode is intentionally limited to OpenAI Priority-priced models:

• gpt-5.6-luna
• gpt-5.6-sol
• gpt-5.6-terra
• gpt-5.5
• gpt-5.4
• gpt-5.4-mini

Supported providers:

• openai
• openai-codex

Other models and providers are left untouched.

Pricing warning

The extension shows a one-line warning when enabled: Fast mode turned on for <model> uses about 2x the token price for lower latency.

Note

The extension is off by default and only modifies provider behavior when all of these are true:

1. Fast mode is on.
2. The current provider is openai or openai-codex.
3. The current API is openai-responses or openai-codex-responses.
4. The current model is gpt-5.6-luna, gpt-5.6-sol, gpt-5.6-terra, gpt-5.5, gpt-5.4, or gpt-5.4-mini.

It does not change prompts, tools, or model selection. Non-eligible provider requests fall through to pi's normal OpenAI/OpenAI Codex streamers unchanged.