100xprompt runs on two in-house frontier models - Flash and Pro. Flash is fast and lightweight for everyday steps; Pro brings deeper reasoning for hard refactors and large changes. You decide which one drives each session, switch between them mid-conversation, and give individual agents their own model. This page covers:

​
Overview

Every task you send routes to an active model - Flash or Pro. The active model is whatever you selected last, or your configured default. Each agent can override it. Lightweight background steps run on Flash so they stay fast and cheap.
You never have to think about model setup to get started - sign in to 100xprompt and both models are ready to go. Everything below is for when you want more control: a different model, per-agent tuning, or a higher reasoning effort.

​
The two models

Both models come from 100xprompt’s hosted service, so there’s nothing to install or configure - sign in once and select whichever fits the task.
ModelBest forTypical use
FlashFast, lightweight workQuick edits, naming sessions, short summaries, tight iteration loops.
ProDeep reasoning and large changesHard refactors, multi-file changes, tricky debugging, and planning.
Model availability updates automatically. To pull the latest catalog on demand, run 100xprompt models --refresh.The base catalog comes from models.dev, but the branded Flash/Pro labels are layered on from https://proxy.100xprompt.com/models. This override fetch is fail-open — if the proxy is unreachable, the raw upstream model ids are shown instead.

​
List available models

Every model 100xprompt can reach is listed with the models command. Model identifiers take the form provider/model - 100xprompt/flash and 100xprompt/pro. 
# List every model you can select
100xprompt models

# Include metadata such as cost and capabilities
100xprompt models --verbose
The list is the authoritative source of what you can select right now, so use it to confirm the exact identifier before you pin one in config.

​
Choose and switch the active model

Switch models two ways: interactively in the app, or declaratively in your config so it applies every time.
Open the Switch model picker from within a session. It shows your favorites and recently used models first, so the model you actually work with is always one keystroke away. Pick one and it becomes active for the current conversation immediately - no restart, no lost context.
1

Open the model picker

Trigger the Switch model action while in a session.
2

Pick a model

Choose Flash or Pro. Selection takes effect instantly.
3

Keep working

Your conversation continues on the new model. You can also cycle quickly between your favorites without opening the full picker.
For a single run from the terminal, pass the model directly:
100xprompt run --model 100xprompt/pro "Refactor the billing module"

​
Route lightweight steps to Flash

Some steps don’t need Pro - generating a session title or a short summary, for example. Point these at Flash with the small_model key. Your primary work stays on the model you chose. 
{
  "model": "100xprompt/pro",
  "small_model": "100xprompt/flash"
}
The small model handles background helper tasks only. Your actual coding work always runs on the active model you selected.

​
Tune reasoning effort with variants

Each model can run at a different reasoning effort. A higher-effort variant thinks longer for harder problems; a lighter one answers faster. Which variants a model supports shows up per model in 100xprompt models. For a single run, pass one with --variant: 
# Run at a higher reasoning effort for a hard problem
100xprompt run --variant high "Find the root cause of this deadlock"
Common variant names are high, max, and minimal, but the exact set depends on the model - check 100xprompt models --verbose to see what a given model supports.

​
Per-agent model choice

Suppose you want your build agent on Pro and your explore agent on Flash. Agents can each run on their own model, automatically, without you switching by hand. Set a model on any agent in your config. 
{
  "model": "100xprompt/pro",
  "agent": {
    "build": {
      "model": "100xprompt/pro"
    },
    "explore": {
      "model": "100xprompt/flash"
    }
  }
}
When an agent has its own model, it always uses that. Any agent without one falls back to the active model. See Agents for the full picture on defining and using agents.
You can also pick an agent (and therefore its model) for a single run with 100xprompt run --agent <name>.

Hosted Models

Sign in to 100xprompt and get Flash and Pro with a single key.

Agents

Give each agent its own model, tools, and role.

Settings

Configure defaults, the small model, and more in 100xprompt.json.