Fast mode

A high-speed inference mode for Theus. Toggle it with /fast; it trades extra usage cost for lower latency on the main loop model.

Fast mode is a research preview. It is billed as extra usage at a premium rate and has its own separate rate limits, independent of your normal session limits.

What it does

When fast mode is on, Theus runs the agent loop against a fast-capable model at higher speed. The mode is a per-user preference stored in your settings; it persists across sessions until you turn it off (or an organization policy or usage limit disables it).

The fast-capable model is Opus 4.6. If your current main-loop model does not support fast mode, turning fast mode on automatically switches the session to the fast-capable model. If your model already supports it, no model switch happens.

The `/fast` command

Run the slash command inside a Theus session:

/fast          # open the toggle dialog
/fast on       # turn fast mode ON directly
/fast off      # turn fast mode OFF directly

With no argument, /fast opens a confirmation dialog. With on or off it applies the change immediately and prints a status line, for example Fast mode ON (adding · model set to Opus 4.6 if a model switch was required) or Fast mode OFF.

The dialog

The picker shows the current state (ON/OFF) plus the pricing for the fast-capable model. Controls:

Tab — toggle between ON and OFF
Enter — confirm the selection
Esc — cancel and keep the previous state

Availability requirements

Before allowing a toggle, Theus checks whether fast mode is available for you. It can be unavailable for several reasons:

Situation	What you see
Free / evaluation account	Fast mode requires a paid subscription (or, for API-key auth, credits to purchase)
Disabled by your organization	Fast mode has been disabled by your organization
Extra usage billing not enabled	Fast mode requires extra usage billing · run `/extra-usage` to enable
Non–first-party provider	Fast mode is not available on Bedrock, Vertex, or Foundry
Network error reaching the status endpoint	Fast mode unavailable due to network connectivity issues

Fast mode is only available with the first-party Theus provider. It is also not available in non-interactive Agent SDK sessions unless explicitly opted in via settings.

Cooldown

Because fast mode has separate rate limits, it can enter a temporary cooldown. This happens when you hit your fast limit (rate limit) or when fast capacity is temporarily overloaded. During cooldown the dialog shows the reason and a countdown to when it resets; fast mode resumes automatically once the reset time passes.

Environment variables

Variable	Effect
`THEUS_CODE_DISABLE_FAST_MODE`	When truthy, disables the fast mode feature entirely. The `/fast` command becomes unavailable and hidden.
`THEUS_CODE_SKIP_FAST_MODE_NETWORK_ERRORS`	When truthy, bypasses the organization availability check when it fails due to network/proxy errors. Useful behind corporate proxies that block the status endpoint.

Related settings

fastMode — the persisted user preference (on/off). Set implicitly by /fast.
fastModePerSessionOptIn — when set, fast mode starts off at the beginning of every session even if the preference is on, requiring a fresh opt-in each time.

Fast mode consumes extra usage at a premium rate. If your organization or plan has extra usage disabled or your credits are exhausted, fast mode will be turned off automatically and you will see a reason-specific message.

Theus — local-first multimodel coding agent · Made in Peru