macrostack
Browse

The AI stack

The whole five-layer mapL3InfrastructureL4Models & toolingL5Applications

Categories

All 78 categories, by layer →

About

How we rank & score

Is OpenAI API (ChatGPT) free? What it actually costs in 2026

The short answer

The OpenAI API has no free tier: every request is billed per million tokens, from $0.05 input on the smallest nano model to $10 input / $50 output on the flagship. Batch halves it; cached input is 90% off.

The pricing page lists no free tier and no free credits. ChatGPT the product has a free plan; the API you build on does not — usage is prepaid or billed monthly per token.

What the free tier actually includes

  • No free tier, no free credits stated on the pricing page
  • Billing is per million tokens, separately for input and output
  • Cheapest listed model: gpt-5-nano at $0.05 input / $0.40 output per million tokens

Where the wall is

There is no wall to hit because there is no free floor to stand on. The real cost surprise is output tokens, priced five times input on the flagship — long answers, not long prompts, are what run up a bill.

After that: gpt-6-luna $0.10 / $0.50 per million tokens (input / output); gpt-6-sol $2 / $10; gpt-6-astra $10 / $50. Mini and nano models run $0.05–$1.10 input. Batch API is 50% off; cached input is 90% off.

Can you self-host OpenAI API (ChatGPT)? No — cloud only

OpenAI's models run only on OpenAI's servers. The alternatives on this page — Ollama, LocalAI, vLLM — run open-weight models on your own hardware with an OpenAI-compatible API, which is why 'drop-in replacement' is the search people make.

What it actually takes to run one yourself

Real RAM against the documented minimum, monthly cost, setup time, and the one thing that catches people out.

Self-hosting Ollama

needs 8 GB VRAM for a 7B model at usable speed; 24 GB for 30B-class (docs say 8 GB system RAM)$0 on hardware you own. $150–400/mo for a rented 24 GB GPU, against per-token API billing that is cheaper below roughly 2M tokens a month.

Self-hosting vLLM

needs 24 GB VRAM minimum for useful production serving (docs say 16 GB system RAM)$300–900/mo for a rented A100 or L40S, against managed inference with a platform margin on every token

Self-hosting LocalAI

needs 8 GB RAM for CPU inference; 8 GB VRAM for anything comfortable (docs say 4 GB)$0 on hardware you own; a rented GPU is $150–400/mo

Self-hosting LM Studio

needs 16 GB system RAM, and VRAM is what actually decides your model (docs say 8 GB)$0 — it runs on the machine you already own

When paying is still the right call

If you need frontier-model quality (top-tier reasoning and the newest capabilities), want zero infrastructure to run, or your volume is low, OpenAI's hosted API is genuinely hard to beat. Self-hosting pays off once you have the GPU, care about data privacy or offline use, or your token bills have outgrown the cost of your own hardware.

What you are locked into

Proprietary models and API surface; your data and cost structure sit entirely on OpenAI's terms.

If you would rather not pay: Ollama

92

Ollama

Run Llama, Mistral, Qwen and more with one command.

Free / self-host (you pay only for your own hardware + power)

See all 8 OpenAI API (ChatGPT) alternatives compared

Compare the free options head-to-head

Ollama vs LocalAI→Ollama vs vLLM→Ollama vs Mistral AI→

Read from OpenAI's own pricing and documentation on 2026-09-23 — 7 days ago. Vendors change prices without notice; confirm before you commit.

The Macrostack brief

New swaps, worth your inbox.

A short, occasional email when we add a high-intent alternative or ship a new head-to-head. No spam, no selling your address — unsubscribe in one click.