AI & Developer Tools

AI Model API Pricing

List prices for all 17 major models from 3 providers, August 2026. The spread is wider than most people assume: on a real chatbot workload, Claude Fable 5 costs 105× what Gemini 2.5 Flash-Lite does — $1.52 a month against $160.

Every tracked model

ModelProviderTierInput $/1MOutput $/1M
Gemini 2.5 Flash-LiteGooglebudget$0.10$0.40
GPT-5.6 LunaOpenAIbudget$0.20$1.20
GPT-5.4 NanoOpenAIbudget$0.20$1.25
Gemini 2.5 FlashGooglebudget$0.30$2.50
Claude Haiku 4.5Anthropicfast$1.00$5.00
Gemini 3.6 FlashGooglefast$1.50$7.50
Gemini 3.7 FlashGooglefast$1.50$7.50
Gemini 3.5 FlashGooglefast$1.50$9.00
Claude Sonnet 5Anthropicflagship$2.00$10.00
Gemini 3.1 ProGoogleflagship$2.00$12.00
GPT-5.6 TerraOpenAIflagship$2.00$12.00
GPT-5.4OpenAIflagship$2.50$15.00
Claude Opus 4.8Anthropicfrontier$5.00$25.00
Claude Opus 5Anthropicfrontier$5.00$25.00
GPT-5.5OpenAIfrontier$5.00$30.00
GPT-5.6 SolOpenAIfrontier$5.00$30.00
Claude Fable 5Anthropicfrontier$10.00$50.00

Ordered by output price, which is what dominates most production bills. Input and output rates do not move together — the ratio between them runs from 4.00× to 8.33× across this table — so a model that looks cheap on input can be the expensive choice for generation-heavy work.

Compare within a provider

Each provider page prices the same six real jobs across that provider's whole lineup, works out what each step up the ladder costs on a live workload, and shows where the provider stands against the other two at every tier it competes in.

Or go straight to the numbers

Standard-tier, short-context list prices, last checked against each provider's own published pricing page on . Long-context requests, priority tiers, batch processing and fine-tuned variants are priced differently. Confirm against the provider's own pricing page before committing budget.

You might also need

Last reviewed: · Who maintains this · How it is checked

Prices are read from each provider's own published pricing page, not from third-party summaries. A check that runs on every build (check-prices.js) fails the deploy if any two pages on this site quote a model differently.