Free tool · Updated 2026-09-30

LLM Pricing Comparison

The cheapest LLM API depends on your traffic, not the sticker. Enter your mix and see the ranking at your own numbers.

Input tokens / request

Everything you send: prompt, context, retrieved documents.

…of which repeats every call

System prompt, tools, fixed docs. 0 if you do not cache.

Output tokens / request

Output bills 3x to 8x input. It moves the ranking.

Requests / month

Sets how often a cache is warm.

Cheapest at your traffic shape

Gemini 2.5 Flash-Lite $50.00/mo

The sticker-cheapest model holds the top spot at your mix, but 2 of 13 models change position once your traffic is priced in. Cheapest is not best: check quality on your own examples before switching.

Model	$/M in	$/M out	Cache	Your $/mo
Gemini 2.5 Flash-Lite Google	$0.10	$0.40	90% off reads	$50.00
GPT-5.6 Luna OpenAI	$0.20	$1.20	90% off reads	$120.00
DeepSeek V4.1 Flash DeepSeek	$0.30	$1.20	98% off reads	$150.00
Gemini 3.8 Flash Google	$0.75	$3.75	90% off reads	$412.50
Claude Haiku 4.5 Anthropic	$1.00	$5.00	90% off reads	$550.00
DeepSeek V4 Pro DeepSeeksticker #7 → yours #6	$1.32	$3.96	96.7% off reads	$594.00
Gemini 2.5 Pro Googlesticker #6 → yours #7	$1.25	$10.00	90% off reads	$875.00
Claude Sonnet 5.5 Anthropic	$2.00	$10.00	90% off reads	$1,100
GPT-5.6 Terra OpenAI	$2.00	$12.00	90% off reads	$1,200
Gemini 3.1 Pro (preview) Google	$2.00	$12.00	90% off reads	$1,200
Claude Opus 5.5 Anthropic	$4.00	$20.00	95% off reads	$2,200
GPT-5.6 Sol OpenAI	$4.00	$20.00	90% off reads	$2,200
Claude Fable 5.1 Anthropic	$10.00	$50.00	97.5% off reads	$5,500

List prices checked 30 Sep 2026 on each provider's pricing page (model names link to it). Standard rates, not batch. DeepSeek shows its peak-hour price; off-peak is half. Cache hits assume requests arrive at random through the month and include each provider's write fee.

Why sticker tables pick the wrong model

Three things a $/MTok column cannot tell you.

Output moves the ranking

Output bills 3x to 8x input and the multiple differs per model. An output-heavy workload reorders the whole table.

Cache terms differ

Reads cost 2% to 10% of input, but write fees, minimum prompt sizes and cache lifetimes differ. A 2,500-token prompt caches on some models and not at all on others.

Sticker is one of three numbers

Input price, output multiple, cache terms. Comparing on the first alone is how teams pick the model that costs more.

The full argument is in how to cut your OpenAI or Anthropic bill, and the wider sequence is in LLM cost optimization.

LLM pricing FAQ

What is the cheapest LLM API?+

On sticker price, Gemini 2.5 Flash-Lite, at $0.10 per million input tokens (September 2026). On your actual bill, it depends on your traffic shape: output bills 3x to 8x input depending on the model, and caching terms differ in write fees, minimum prompt size and how long a cache lives, so the cheapest model at one mix is regularly not the cheapest at another. The comparison above ranks by effective monthly cost at your own numbers.

Why is the cheapest model per token not the cheapest overall?+

Three reasons. Output multiples differ, so output-heavy work reorders the table. Cache discounts differ, so a stable prefix at high traffic can make a dearer model cheaper in practice. And failure rates differ: a cheap model that retries pays for every failed attempt, which is invisible in a price table and dominant in a real bill.

How do I compare LLM API pricing properly?+

Price a month of your real traffic, not a million tokens. Take your input and output tokens per request, how much of the prompt repeats, and your request volume, and compute the monthly figure per model. Then check quality on your own evaluation set, because the cheapest model that fails your task is the most expensive option on the list.

What does LLM inference actually cost?+

Inference cost is price times volume across three token types: input, output at 3x to 8x the input rate, and on reasoning models the internal thinking tokens, which bill as output without appearing in the response. The comparison above prices your mix directly; the per-token rate alone is the least informative of the three numbers.

How often do LLM prices change?+

Several times a year per provider, usually downward, and occasionally with structural changes like new cache tiers that move effective cost more than the sticker change does. This page shows its verification date and links each provider directly, so you can confirm before committing an architecture to a number.

Should I switch providers to save money?+

Rarely as a first move. Provider choice spans roughly 2x on a comparable tier, while fixing caching and routing on the provider you already use spans 3x or more, without a migration or re-evaluating every prompt against a new model. Switch for capability, latency or rate limits; optimise for cost.

Free consultation

Priced the models. Now price the architecture.

The table shows what tokens cost. What you pay is decided by loops, caching and routing. Book a free cost review.

Book a Free 30-Minute Call

We map one workflow that eats your team’s week and show you what an engineer and a few AI agents could take off it. No slide deck, no sales pitch. Just a working session.

Don’t want a call? Email [email protected]

Book a free call

One workflow mapped with you, live
A clear first step for your team
Reply within one working day

We worked with Seif on the React Native app for our vehicle-appraisal platform. What stood out most was how much ownership he takes: he thinks a feature through, raises edge cases before they turn into bugs, and delivers something that actually works. And then there is one more thing why I wanted to work with Seif after our very first call: Exceptional clear and consistent communication. It is a pleasure to work with him and his team!