What is prompt caching and how does it reduce API bills?

Modern frontier providers like Anthropic, OpenAI, and Google support prompt caching. When substantial system prompts, schemas, or reference documents are reused across requests, input tokens from the cached prefix are billed at an 80% to 90% discount, drastically lowering cost in conversational agents and RAG pipelines.

How frequently are these LLM API rates updated?

Rates are maintained in sync with official provider pricing sheets as models are released and prices cut. The calculator also supports custom per-million token overrides if your organization receives enterprise volume discounts.

Are my pasted prompts uploaded or recorded?

No. All token calculations and character estimations occur 100% locally inside your browser. No prompts, queries, or projected usage metrics are sent to any server.

AI Token & LLM API Cost Calculator

Compare OpenAI, Claude, Gemini, and DeepSeek API costs. Estimate prompt tokens, completions, prompt caching savings, and monthly spend.

Architecture Presets:

Cost per Request

$0.0077

1,850 total tokens/call

Daily Spend

$15.45

2,000 requests / 24h

Monthly Projection

$469.94

~61k monthly queries

Annual Run-Rate

$5,639

Assuming steady load

Primary LLM Provider & Model (30 Available):

Prompt Tokens (Input / Request)1,500 tokens

50 (Short)4,000 (Standard)16,000 (RAG)32,000 (Large Doc)

Completion Tokens (Output / Request)350 tokens

20 (Concise)500 (1 page)2,000 (Essay)8,000 (Max gen)

Daily Requests2,000 queries / day

50 (Testing)1,000 (Early MVP)25,000 (Scale)100,000 (Enterprise)

Prompt Cache Hit Ratio50% cached

Anthropic applies a 90% price discount on cached prompt tokens.

Prompt Token Estimator~4 chars/token

Claude 3.7 Sonnet Official Rates:Anthropic

Input (Uncached):$3.000 / 1M tokens

Input (Cached):$0.3000 / 1M tokens

Output / Completion:$15.000 / 1M tokens

Context Window:200k tokens

Cross-Provider Cost Comparison (30 Models)

Ranked comparison based on your current inputs (1500 in, 350 out, 2000 req/day)

Sort:

Model / Provider	Rates (In / Out per 1M)	Cost / Request	Monthly Spend	Variance
Gemini 1.5 Flash-8B(Google)	$0.0375 / $0.15	$0.0001	$5.33	Save $464.61/mo
Llama 3.1 8B(Meta)	$0.05 / $0.08	$0.0001	$6.27	Save $463.68/mo
Amazon Nova Lite(Amazon)	$0.06 / $0.24	$0.0001	$8.53	Save $461.41/mo
Gemini 1.5 Flash(Google)	$0.075 / $0.3	$0.0002	$10.66	Save $459.28/mo
DeepSeek V3(DeepSeek)	$0.14 / $0.28	$0.0002	$12.99	Save $456.95/mo
Gemini 2.0 Flash(Google)	$0.1 / $0.4	$0.0002	$14.22	Save $455.72/mo
Gemini 2.0 Flash Thinking(Google)Reasoning	$0.1 / $0.4	$0.0002	$14.22	Save $455.72/mo
GPT-4o mini(OpenAI)	$0.15 / $0.6	$0.0004	$23.04	Save $446.90/mo
Mistral Small 3 (24B)(Mistral)	$0.2 / $0.6	$0.0005	$31.03	Save $438.92/mo
QwQ 32B (Reasoning)(Qwen)Reasoning	$0.3 / $0.3	$0.0006	$33.76	Save $436.18/mo
Grok 2 mini(xAI)	$0.2 / $1	$0.0006	$39.54	Save $430.40/mo
Qwen 2.5 72B(Qwen)	$0.35 / $0.4	$0.0007	$40.45	Save $429.49/mo
Llama 3.3 70B(Meta)	$0.4 / $0.4	$0.0007	$45.02	Save $424.93/mo
Codestral (2501)(Mistral)Coding	$0.3 / $0.9	$0.0008	$46.54	Save $423.40/mo
DeepSeek R1(DeepSeek)Reasoning	$0.55 / $2.19	$0.0013	$78.00	Save $391.95/mo
Amazon Nova Pro(Amazon)	$0.8 / $3.2	$0.0019	$113.76	Save $356.18/mo
Claude 3.5 Haiku(Anthropic)	$0.8 / $4	$0.0021	$125.32	Save $344.62/mo
o3-mini(OpenAI)Reasoning	$1.1 / $4.4	$0.0028	$168.97	Save $300.98/mo
o1-mini(OpenAI)	$1.1 / $4.4	$0.0028	$168.97	Save $300.98/mo
Gemini 2.0 Pro (Exp)(Google)	$1.25 / $5	$0.0029	$177.75	Save $292.19/mo
Gemini 1.5 Pro(Google)	$1.25 / $5	$0.0029	$177.75	Save $292.19/mo
Llama 3.1 405B(Meta)	$1.8 / $1.8	$0.0033	$202.58	Save $267.37/mo
Mistral Large 2(Mistral)	$2 / $6	$0.0051	$310.25	Save $159.69/mo
GPT-4o(OpenAI)	$2.5 / $10	$0.0063	$384.01	Save $85.93/mo
Grok 2(xAI)	$2 / $10	$0.0065	$395.42	Save $74.52/mo
Claude 3.7 Sonnet(Anthropic)Hybrid ReasoningSelected	$3 / $15	$0.0077	$469.94	—
Claude 3.5 Sonnet(Anthropic)	$3 / $15	$0.0077	$469.94	+$0.00/mo
o1(OpenAI)Reasoning	$15 / $60	$0.0379	$2,304.09	+$1,834.15/mo
Claude 3 Opus(Anthropic)	$15 / $75	$0.0386	$2,349.71	+$1,879.77/mo
GPT-4.5 (Preview)(OpenAI)	$75 / $150	$0.1369	$8,326.65	+$7,856.71/mo

Capabilities

Key Features

Comprehensive pricing catalog covering OpenAI, Claude, Gemini, and DeepSeek
Instant cost-per-call, daily spend, monthly projection, and annual run-rate metrics
Prompt caching toggle to model 50% to 90% savings on cached context and system prompts
Side-by-side model comparison table sorting providers from lowest to highest monthly cost
Interactive sample prompt token estimator approximating tokens from raw text
Preconfigured industry scenarios: Customer Support, Document RAG, and Code Generation
100% private in-browser computation with zero tracking or telemetry
Custom token pricing override for fine-tuned or enterprise negotiated rates

Quick Guide

How to use

Select an AI model or preset

Choose a primary language model from OpenAI, Anthropic, Google, or DeepSeek, or pick a pre-configured architecture preset.

Configure token volume

Adjust sliders for input prompt tokens and completion tokens per request, or paste your prompt to auto-estimate token length.

Set query request volume

Define your expected daily or monthly API request frequency to simulate production traffic loads.

Compare providers and optimize

Inspect the sorted comparison table to identify potential savings by switching models or leveraging prompt caching.

Technical Guide

About AI Token & LLM API Cost Calculator

AI Token & LLM API Cost Calculator is an interactive financial planning utility for AI engineers, founders, and developers building with large language models. Seamlessly compare API expenses across OpenAI (GPT-4o, o1, o3-mini), Anthropic (Claude 3.7 Sonnet, Claude 3.5), Google (Gemini 2.0 Flash, 1.5 Pro), DeepSeek (V3, R1), and Mistral. Calculate cost-per-request, daily burn rate, and monthly projections based on prompt tokens, completion volume, and prompt caching discounts.

FAQ

Frequently Asked Questions

In standard English text, 1 token represents roughly 4 characters or approximately 0.75 words. A 1,000-word essay translates to roughly 1,333 tokens. Non-English languages, specialized code syntax, and JSON payloads typically require more tokens per word due to multi-byte subword tokenization.

Related Tools

Browse all

AI Token & LLM API Cost Calculator

Cross-Provider Cost Comparison (30 Models)

Key Features

How to use

Select an AI model or preset

Configure token volume

Set query request volume

Compare providers and optimize

About AI Token & LLM API Cost Calculator

Frequently Asked Questions

Related Tools

CSS Clamp Fluid Typography Calculator

Can You Defend This Cricket Total?

Smart Text Diff Checker

Cross-Provider Cost Comparison (30 Models)