pi-codex-fast

Fast-mode extension for pi that injects service_tier: "priority" into supported OpenAI Codex requests.

Usage

Inside pi:

• /codex-fast to toggle

From CLI:

• pi --fast

Persistence

The enabled/disabled state is read from pi's settings files:

• global: $PI_CODING_AGENT_DIR/settings.json (or ~/.pi/agent/settings.json)
• project override: <cwd>/.pi/settings.json

under the key pi-codex-fast.enabled.

Writes go to the global settings file.

Behavior

The extension only patches provider payloads when all of these are true:

• fast mode is enabled
• the active model is one of:
  • openai-codex/gpt-5.4
  • openai-codex/gpt-5.5
  • openai-codex/gpt-5.6-sol
  • openai-codex/gpt-5.6-terra
  • openai-codex/gpt-5.6-luna

All other requests are left unchanged.

Example benchmark

A local live benchmark is available in this repository under evals/. Three paired trials per model produced:

Model	TTFB speedup	Turn speedup	Wall speedup
gpt-5.6-sol	1.52x	1.58x	1.57x
gpt-5.6-terra	1.05x	1.34x	1.33x
gpt-5.6-luna	1.30x	2.31x	2.25x