Context Modes

Context modes let you control how much of a conversation Pentos sends to the AI on each response. Sending more history helps the AI recall earlier details, but it also costs more and uses more energy. Pentos gives you three modes so you can pick the right balance for each task: Lean, Standard, and Max.

In every mode, your full conversation is always preserved on your account. The context mode only changes how much of that history is included with your next message.

The Three Modes

• Lean (default): The AI receives only your last 10 messages. This is the lowest cost, the fastest, and the most energy efficient option, and it is our recommended default for focused, task-by-task work. Lean adds no points to a response.
• Standard: The AI packs as much of your recent history as fits inside the model’s standard context window, dropping the oldest message pairs first. This keeps responses at each model’s standard pricing in nearly all cases (see Cost and Long-Context Pricing). Each response in Standard costs 1 more point.
• Max: The AI uses the model’s full maximum context window, over 1,000,000 tokens on some models, so very long conversations and large documents stay in view. Max Mode needs a paid plan, and each response in Max costs 3 more points.

For many models the standard window already is the full window, so Standard and Max behave identically. The difference shows up on models with an extended long-context window. For example, Grok 4.7 has a 200,000-token standard window and up to 500,000 tokens in Max, and GPT 6 Luna has 272,000 tokens standard and about 1,050,000 in Max.

If a conversation is too long for the selected model even after trimming, Pentos tells you before sending and suggests Max Mode or a model with a larger window.

Choosing a Mode Per Message

Open the chat input menu (the + button at the bottom left of the chat input area), hover Context, and pick Lean Mode, Standard Mode, or Max Mode. Each mode that adds points shows how many next to its name. The choice applies to your next message and stays in place for the rest of the chat until you change it.

On Free, picking Max Mode opens the upgrade options. A Max default saved while you were on a paid plan is used as Standard until you upgrade again.

Setting a Default

To set the mode every new chat starts in, open Settings, choose Chats from the left sidebar, go to the Preferences tab, and under AI Settings open the Context dropdown. Pick Lean, Standard, or Max (Max needs a paid plan). The choice saves immediately, and every new chat starts in your chosen mode.

Pairing With the Pentos Smart Router

Context modes and the Pentos Smart Router are designed to work together. The Smart Router picks the right-sized model for each message, while your context mode controls how much history that model receives. Lean paired with the Smart Router keeps both the model and the context small, which is where the cost and energy savings come from, and is why we recommend both as your default. See Energy Usage for how the savings are estimated and what they assume.

Cost and Long-Context Pricing

Most providers charge a higher per-token rate once a request’s context grows beyond the model’s standard window. Pentos reflects this honestly: when a request crosses that point, the whole request is estimated at the higher rate, so your per-message cost and energy estimates stay accurate.

Because Lean and Standard size the conversation to the standard window, this surcharge almost always applies only in Max Mode, and only on models that actually have a higher long-context tier. Pentos sizes your history with an estimate, and some models count the same text as more tokens, so a response that starts near the limit, or one that uses many tools (each search, connector action, or file read adds to what the model reads next), can still pass the standard window partway through. This happens most often on Claude Haiku 5.5, whose standard window is 100,000 tokens. The surcharge never changes the points a response costs, but on your own API key your provider bills it. Some models, including Claude Opus 5.5 and Claude Sonnet 5.5, offer their full window at standard pricing, so no surcharge applies to them in Max.

Energy Transparency

The savings from Lean Mode grow with the length of your conversation. On a short chat there is little history to trim, so the effect is small. On a long thread, where the full history would otherwise be resent with every response, Lean Mode can help reduce the estimated energy needed by reducing the size of the context it has to evaluate.

Because Lean Mode sends only the most recent messages, the AI may not recall something from earlier in the chat unless it was saved as a Memory and Personalization is enabled. We recommend Lean Mode for focused, task-by-task work, Standard for longer evolving conversations, and Max when a long thread or a large document needs to stay fully in context.

Curious how much you are saving? Pentos shows estimated energy consumption per message and per chat. See Energy Usage for the full breakdown by conversation length, how those estimates are calculated, and what each model consumes.