StackProofStackProof

LLM cost, measured

What does an LLM task actually cost?

Current LLM API prices from official provider pages, plus interactive calculators that show what an agent task really costs and when self-hosting beats the API. Open data, open math.

cost / task · 3-call agent task live
  1. Mistral Small$0.00158
  2. DeepSeek-V4-Flash$0.00396
  3. Gemini 2.5 Flash$0.00510
  4. Grok 4.3$0.00938
  5. GPT-5.4 mini$0.010
  6. Claude Haiku 4.5$0.012
  7. Claude Sonnet 5$0.024
  8. Gemini 3.1 Pro$0.027
  9. Claude Opus 4.8$0.060
  10. GPT-5.5$0.068
cheapest
$0.00158
priciest
$0.068
spread
43x
01

Current prices

Standard input/output rates from official provider pages, not memory or scraped listicles.

02

AIO-resistant

Interactive $/task tools an AI Overview can't compute for you, so the click stays here.

03

Open + reproducible

The pricing dataset is downloadable CC-BY, and the cost math is open and re-runnable.