The Chatbot: The Only Thing on Gendia Priced by the Word

Everything else on Gendia charges you a fixed amount for a run. The chatbot charges you for words, yours going in, and its coming out.

That's the one thing to understand before using it, because it makes the cost behave in a way nothing else here does.

---

The rates

Model	Input	Output
GPT-5 Mini	0.2 credits / 1K tokens	1.3 credits / 1K tokens
GPT-5	0.8 credits / 1K tokens	6.5 credits / 1K tokens

Minimum charge of 1 credit per message.

Two things jump out of that table.

Output costs about six times what input does. So a long question with a short answer is cheap; a short question with a long answer is not. "Write me a 2,000-word article" is the expensive shape of request, and "here are three pages of notes, give me a one-line summary" is the cheap one.

GPT-5 costs four to five times GPT-5 Mini. For drafting, summarising, brainstorming and rewriting, which is most of what anyone uses a chatbot for, Mini is genuinely fine. Save the full model for reasoning that Mini visibly fumbles.

A rough sense of scale: 1,000 tokens is around 750 words. A normal back-and-forth exchange on Mini costs the 1-credit minimum. You'd have to work at it to spend what a single image costs.

---

The two toggles

Reasoning effort ("Thinking"). Higher settings let the model think longer before answering, and those thinking tokens are output tokens, which are the expensive ones. Raise it for genuinely hard problems; leave it low for everything else. This is the setting most likely to quietly multiply a bill.

Web search. Lets the model look things up rather than rely on what it already knows. Turn it on for anything current, prices, recent releases, what a competitor is doing now, and off otherwise, since fetched pages become input tokens.

---

What it's actually good for here

The chatbot is most valuable not as a thing in itself but as the cheap step before the expensive ones:

Writing prompts. Describe what you want in plain words and ask it to write you an image or video prompt with subject, setting, light, framing and camera move. A 1-credit exchange that improves a 125-credit generation is the best trade on the platform.

Planning a video. Ask it to break your idea into 5-second shots. You get a shot list before you spend anything on video.

Ad copy and scripts. Headlines, product descriptions, the narration you'll run through text to speech.

Naming and angles. Twenty product-name options for about a credit.

Deciding what to make before you commit credits to making it.

---

Where it fits with everything else

There's an obvious loop, and it's worth using deliberately:

1. Chatbot, work out what you're making and write the prompt. Costs ~1 credit.
2. Flux Schnell, test the prompt. 2.5 credits.
3. The real model, generate it properly. 20 to 200 credits.

The two cheap steps cost less than 4 credits together and dramatically improve the third. Most people skip straight to step 3 and iterate there, which is the most expensive place to think.

---

Keeping it cheap

• Default to Mini. Switch up only when it's clearly struggling.
• Keep reasoning low unless the problem is hard.
• Web search off unless you need something current.
• Ask for what you need, not more. "Give me five headlines" costs a fraction of "write a full campaign brief", and you probably wanted the headlines.
• Start a new conversation for a new topic. Long threads resend their history as input tokens on every message, so an unrelated question at the end of a long chat carries the whole conversation's weight.

That last one is the one nobody notices.

---

Try it

Open the chatbot on GPT-5 Mini and ask it to write you an image prompt for something you want to make, with subject, setting, light, framing and lens. Then run that prompt on Flux Schnell. Under 4 credits for the pair, and it's the workflow most worth building a habit around.

Then: prompt writing for images · what everything costs