AI Batch: wait a few hours, pay less

Not every request to the AI is equally urgent. Some you make while writing and need answered now; others you could launch before closing the laptop and find done the next day. NovLore treats these two cases differently, and knowing which is which saves credits.

What AI Batch is

AI Batch puts your job in a queue instead of running it instantly. The answer arrives after hours, not seconds. In exchange for the wait, the processing costs less: the API NovLore uses for these jobs — Claude's Batch API — is priced at roughly half the same request in real time.

It isn't a worse version of the model. It's the same model, with delayed delivery.

When the wait is a good deal

Batch pays off when the task is heavy and you aren't watching it:

• generating a whole chapter from the outline and the story bible;
• a broad pass over the book's structure;
• bulk jobs you'd plan to leave running while you do something else.

In all of these, sitting in front of the screen with a spinner gave you no advantage. You might as well move the work to the queue and pay less for it.

When you do need the answer now

Real time stays the right choice when latency is exactly the point:

• chat with Spark, where you're thinking alongside the assistant;
• editing a selected passage, which you make and review within seconds;
• on-the-fly correction while you write.

Here a wait of hours would break the workflow instead of helping it. Full cost is the price of promptness, and in these cases promptness is worth it.

How it works in NovLore

You send the job to the queue, the app records it and lets you keep writing. It checks the status at intervals; when the result is ready it's waiting for you. You don't have to keep the page open or the computer on.

The practical rule is a single one: if you don't need to watch the AI work, send it to batch. If you're using it to think, keep it real time.

Frequently asked questions

How long do I have to wait, exactly?

It depends on load. Usually it's hours, not days. It's meant for work you launch and pick up later, not for something you need within the same session.

Is the batch result lower quality?

No. The model is the same and the instructions are the same: only the delivery time changes. The only trade-off is time, not text.