Compare LLM API Pricing

19 models across 5 providers, sortable by price. Click a source link to verify any figure directly.

ProviderModelInput $/M ↑Output $/MVerifiedConfidenceSource
OpenAIGPT-5 nano$0.05$0.42026-10-25Officiallink
AnthropicHaiku 5.5
Rate for prompts up to 100K tokens; longer prompts cost $0.50 in / $2.50 out.
$0.1$0.52026-10-25Officiallink
OpenAIGPT-6 Luna$0.1$0.52026-10-25Officiallink
GoogleGemini 3.1 Flash-Lite$0.25$1.52026-10-25Officiallink
GoogleGemini 3.5 Flash-Lite$0.3$2.52026-10-25Officiallink
DeepSeekDeepSeek V4.1 Flash
Peak rate (weekdays 01-04 and 06-10 UTC). All other hours cost half: $0.15 in / $0.60 out.
$0.3$1.22026-10-25Officiallink
GoogleGemini 3.8 Flash
Price doubles on 1 Jan 2027 ($1.50 in / $7.50 out).
$0.75$3.752026-10-25Officiallink
xAIGrok Build 0.1
Prompts of 200K tokens or more cost double.
$1$22026-10-25Officiallink
DeepSeekDeepSeek V4 Pro
Peak rate (weekdays 01-04 and 06-10 UTC). All other hours cost half: $0.66 in / $1.98 out.
$1.32$3.962026-10-25Officiallink
AnthropicSonnet 5.5$2$102026-10-25Officiallink
OpenAIGPT-6.1 Sol$2$102026-10-25Officiallink
GoogleGemini 4 Argon
Flagship reasoning model. Launch price until 30 Nov 2026; from 1 Dec 2026 it doubles to $4 in / $20 out. Implicit caching only.
$2$102026-10-25Officiallink
GoogleGemini 3.1 Pro
Preview. Prompts over 200K tokens cost $4 in / $18 out.
$2$122026-10-25Officiallink
xAIGrok 4.7
Prompts of 200K tokens or more cost double.
$2$62026-10-25Officiallink
AnthropicOpus 5.5$4$202026-10-25Officiallink
OpenAIGPT-5.6 Sol
Promotional price, guaranteed at least until 21 Nov 2026.
$4$202026-10-25Officiallink
AnthropicFable 5.1
Flagship reasoning model.
$10$502026-10-25Officiallink
OpenAIGPT-6 Astra$10$502026-10-25Officiallink
OpenAIGPT-5.5 Pro
Rate for context under 272K tokens. No cached-input discount.
$30$1802026-10-25Officiallink
Note: Prices marked Secondary source come from a provider's official pricing page that blocked automated verification (common for API docs behind bot protection) — cross-checked across multiple independent sources instead of a single one, but not confirmed directly. Verify before relying on it for a large workload.

How the numbers are calculated

Each row is one model’s standard API price in US dollars per million tokens, taken from the provider’s official pricing page and dated with the day we last checked it. Click a column heading to sort; the source link opens the page the number came from.

Notes under a model flag what a single number cannot show: long-context surcharges, DeepSeek’s peak and off-peak rates, promotional prices and announced increases. Every change we have made to these numbers is listed, with its date, in the price history.

Worked example with current prices

Take a workload of 10 million input tokens and 2 million output tokens a month.

The cheapest model in the table for that mix is OpenAI GPT-5 nano at $1.30 a month; the most expensive is OpenAI GPT-5.5 Pro at $660, about 508 times as much.

Flip the mix to 2 million input and 10 million output tokens and the cheapest is OpenAI GPT-5 nano at $4.10, while OpenAI GPT-5.5 Pro costs $1,860. As output grows, models with a low output price move up the ranking, so compare both price columns against your own mix, not the input price alone.

Common mistakes

When not to rely on this tool

Sources

Every price comes from the provider’s own pricing page:

Last reviewed:

Go deeper: What a token is, and how to estimate tokens for your text · Prompt caching and batch discounts, provider by provider · How to estimate the monthly cost of a chatbot, RAG app or agent · Why output tokens cost more than input tokens · Price history and changelog

Frequently asked questions

Which LLM API is the cheapest?

It changes often. Sort the table by input or output price to see the current cheapest models, and check the verified date and source link for each one.

Should I compare input or output prices?

Both, weighted by your workload. Chat and summarisation send far more input than they get back; generation-heavy tasks like writing are dominated by output cost.

Is the cheapest model good enough?

Often, for simple classification, extraction or short answers. Test a few models on your real prompts — a cheaper model that needs retries can cost more overall.

How current are these prices?

Each row shows when it was last verified and links to its source. Providers change prices frequently, so confirm on the provider's own page before committing to a large workload.

More on QuotaCost

Embed this tool

Free to use on your own site, docs, or blog post — paste this code where you want it to appear.