DEV Community

Manohar Papasani
Manohar Papasani

Posted on

OpenAI's cheapest models per million tokens, and none can be ranked for business work yet.

gpt-5-nano has the lowest rate on OpenAI's pricing page, at $0.05 per 1M input tokens.

Prices are from OpenAI's pricing listings as of August 21, 2026.
No independent test results were found, so no model gets a tier yet.

  • What does each 1M tokens cost you?
  • Is the answer good enough for the job?

⏳ NOT RANKED YET
gpt-5-nano (OpenAI)

  • $0.05 input, $0.005 cached and $0.40 output per 1M tokens.
  • Cheapest rate on the entire OpenAI pricing page. gpt-5-mini (OpenAI)
  • $0.25 input, $0.025 cached and $2.00 output per 1M tokens. gpt-4o-mini (OpenAI)
  • $0.15 input, $0.075 cached and $0.60 output per 1M tokens.
  • Context window of up to 128K tokens. gpt-4o (OpenAI)
  • $2.50 input, $1.25 cached and $10.00 output per 1M tokens.

📌 WHY NO TIER YET
Independent results

  • Only vendor pricing pages were found, so no model is ranked for quality or speed.

What this means for your team:

  1. Compare output prices separately, since output costs 4x to 8x more than input on these models.
  2. Check cached input rates, which are half the normal input price on gpt-4o and gpt-4o-mini.
  3. Wait for independent test results before choosing a model for customer-facing work.

Pick one task this week, run it on two of these models, and compare the cost per answer.

P.S. gpt-4o-mini supports a context window of up to 128K tokens, according to its listing.


Originally shared on LinkedIn.

An AI interview question every day, leadership for tech people and hands-on AI learning:

Top comments (0)