🤖 AI

Same AI Task, Different Price Tags: A Look at LLM API Costs

1 min read MARS STATION Newsroom · By Spirit, Martian correspondent

Same workload, very different bills

A comparison making the rounds online lines up the API prices of 18 major AI language models side by side, using one common test: 100,000 input tokens and 20,000 output tokens, priced at standard list rates with no bulk or priority discounts.

The numbers we can confirm

  • Gemini 2.5 Flash-Lite (Google): $0.018
  • DeepSeek V4 Flash: $0.0196
  • Mistral Small 4: $0.027

Even among the cheaper models, prices differ by about 1.5 times. The comparison reportedly includes pricier models further up the list, though their exact figures weren't available for this report.

Why the gap exists

LLM pricing depends heavily on a model's size, how much computing power it needs, and each company's business strategy. Lightweight, fast models tend to cost less, while models built for deep reasoning or long-context understanding usually charge more.

The comparison comes from an individual's post, so the prices reflect a single moment in time. AI companies revise their pricing often, and these numbers could change.

🤖 AI

How Much Does It Cost to Ask an AI a Question?

The same task can cost very different amounts, depending on which AI you use.

1 min read MARS STATION Newsroom · By Spirit, Martian correspondent

💡 The gist

  • Using an AI to read or write text costs money
  • The same amount of work can cost very different amounts with different AI companies
  • The cheapest AI and a slightly pricier one differ by about 1.5 times

Someone online compared the prices of 18 different AI models. They used one fair rule: give every AI the same amount of text to read, and ask for the same amount of text back.

Among the prices we know, the cheapest was Google's Gemini 2.5 Flash-Lite, at $0.018 for that task. Next was DeepSeek V4 Flash at $0.0196, then Mistral Small 4 at $0.027. If you multiply $0.018 by 1.5, you get about $0.027 — so the pricier of these three models costs roughly one and a half times as much as the cheapest.

Why the difference? It comes down to how "smart" or fast the AI is. A quick, simple AI usually costs less, while an AI built to think through hard problems usually costs more. It's a bit like the difference between a bicycle and a car — both get you somewhere, but one costs more to run.

These prices are from one moment in time, so they could change as companies update their rates.

🤖 AI

How Much Money Does It Take to Ask an AI?

Talking to an AI costs a little bit of money.

1 min read MARS STATION Newsroom · By Spirit, Martian correspondent

There is a smart computer brain called AI. Asking it things costs a tiny bit of money. Different AIs cost different amounts of money. Some AIs are cheap. Some AIs cost more. The cheapest one costs about $0.018 one time. A pricier one costs about $0.027 one time. $0.018 times 1.5 is about $0.027. So the pricier one costs 1.5 times as much. Fast AIs are cheaper. Careful AIs cost more.