LLM Cost Calculator
This LLM cost calculator estimates your API cost per request, per day and per month for current OpenAI, Anthropic Claude and Google Gemini models. Type token counts or paste a prompt to count them automatically.
Reasoning (thinking) tokens are billed as output tokens.
Part of each prompt served from the provider’s prompt cache (for example a long, unchanged system prompt).
| Model | Input / 1M | Cached / 1M | Output / 1M | Per request | Per day | Per month |
|---|
Prices checked: November 2026. Standard (non-batch) API prices in US dollars; a month is counted as 30 days. Always confirm on the provider’s pricing page before you budget.
OpenAIAnthropicGoogle Cheapestmonth input tokensestimateEstimate Your LLM API Bill Before You Ship
LLM APIs charge per token, with separate prices for input, cached input and output, and those prices differ by up to 80 times between the smallest and the largest current models. This LLM cost calculator turns your expected traffic into a cost per request, per day and per month for every current model side by side.
Enter the tokens a typical request sends and receives and your expected volume. The table updates instantly, marks the cheapest option and applies long-context and scheduled price changes. Prices were checked against the official pricing pages in November 2026.
How to calculate LLM API costs
- Measure a typical request. Enter the input tokens (system prompt, conversation history, retrieved documents and the user message), or click Count tokens from a prompt and paste the prompt: each model then gets its own token count. Estimate the output length.
- Enter your volume as requests per day or per month.
- Set the cached share if part of every prompt is identical and you use prompt caching.
- Compare. Filter by provider, sort by monthly cost and copy the table as CSV for a spreadsheet.
The cost formula
Providers quote prices per million tokens (MTok). For one request:
cost = (uncached input × input price + cached input × cached price + output × output price) ÷ 1,000,000
Example: 2,000 input and 500 output tokens on a model priced at $2 input and $10 output per MTok costs 2,000 × 2 ÷ 1M + 500 × 10 ÷ 1M = $0.004 + $0.005 = $0.009 per request. At 1,000 requests a day that is $9 a day or $270 for a 30-day month. If half of the input is a cache hit at $0.10 per MTok, the input part drops to $0.0021 and the month to $213.
LLM API prices (per 1M tokens)
Standard API prices for input, cached input and output, checked in November 2026 on the official pages: OpenAI API pricing, Claude API pricing and Gemini API pricing.
| Model | API ID | Input | Cached input | Output | Notes |
|---|---|---|---|---|---|
| GPT-6.2 Astra | gpt-6.2-astra | $8 | $0.40 | $40 | Prompts over 272K tokens: $16 / $0.80 / $60 |
| GPT-6.2 Sol | gpt-6.2-sol | $2 | $0.10 | $10 | Prompts over 272K tokens: $4 / $0.20 / $15 |
| GPT-6.1 Terra | gpt-6.1-terra | $1 | $0.05 | $5 | Prompts over 272K tokens: $2 / $0.10 / $7.50 |
| GPT-6 Luna | gpt-6-luna | $0.10 | $0.01 | $0.50 | Prompts over 272K tokens: $0.20 / $0.02 / $0.75 |
| GPT-5.5 | gpt-5.5 | $5 | $0.50 | $30 | Prompts over 272K tokens: $10 / $1 / $45 |
| GPT-5.4 mini | gpt-5.4-mini | $0.75 | $0.075 | $4.50 | – |
| Claude Fable 5.2 | claude-fable-5-2 | $8 | $0.20 | $40 | – |
| Claude Opus 5.5 | claude-opus-5-5 | $4 | $0.20 | $20 | – |
| Claude Sonnet 5.5 | claude-sonnet-5-5 | $2 | $0.10 | $10 | – |
| Claude Haiku 5.5 | claude-haiku-5-5 | $0.10 | $0.01 | $0.50 | Prompts over 100K tokens: $0.50 / $0.05 / $2.50 |
| Claude Sonnet 4.6 | claude-sonnet-4-6 | $3 | $0.30 | $15 | – |
| Claude Haiku 4.5 | claude-haiku-4-5 | $1 | $0.10 | $5 | – |
| Gemini 4 Carbon | gemini-4-carbon | $2 | $0.10 | $10 | From 2026-12-01: $4 / $0.20 / $20 (announced) |
| Gemini 4 Argon | gemini-4-argon | $2 | $0.10 | $10 | From 2026-12-01: $4 / $0.20 / $20 (announced) |
| Gemini 4 Neon | gemini-4-neon | $3 | $0.15 | $15 | – |
| Gemini 4 Hydrogen | gemini-4-hydrogen | $1.50 | $0.075 | $7.50 | – |
| Gemini 3.8 Flash | gemini-3.8-flash | $0.75 | $0.075 | $3.75 | From 2027-01-01: $1.50 / $0.15 / $7.50 (announced) |
| Gemini 3.5 Flash-Lite | gemini-3.5-flash-lite | $0.30 | $0.03 | $2.50 | – |
What the calculator includes and leaves out
- Long-context pricing: OpenAI’s GPT-6 family and GPT-5.5 charge higher rates when the prompt is over 272K tokens, and Claude Haiku 5.5 when it is over 100K tokens. The calculator switches to those rates automatically.
- Scheduled changes: Google has announced higher prices for Gemini 4 Carbon and Argon from 2026-12-01 and for Gemini 3.8 Flash from 2027-01-01. The table uses the price valid on the day you open it.
- Cache writes: Anthropic bills writing to the cache at 1.25× the input price (5-minute cache) and Google adds a storage fee per hour. Those one-off costs are not included; the cached share uses the cache-read price.
- Not included: batch discounts, priority or fast modes, regional endpoint premiums, tool fees (such as web search) and taxes.
Ways to lower your LLM costs
- Pick the smallest model that passes your tests; it can cost up to 80 times less than the flagship.
- Cache stable prefixes. Put the system prompt, tool definitions and reference documents first so they can be cached.
- Use batch processing for work that can wait: Anthropic’s Message Batches API is 50% off, and OpenAI and Google also offer discounted batch modes.
- Limit output. Ask for concise answers and structured JSON (build the schema with the JSON Schema Builder), and set a sensible maximum output length.
Frequently Asked Questions
Common questions about LLM API pricing and this calculator.
The math is exact for the token counts you enter, using the standard list prices checked in November 2026. Your real bill depends on how many tokens your requests really use, plus extras such as cache writes, tools and taxes.
Input tokens are processed in parallel, while output tokens are generated one at a time, which uses far more compute per token. For the models in this calculator, output costs five to about eight times as much as input.
Yes. Reasoning models think before they answer, and those tokens are billed as output tokens even if you do not see them. Add a margin to your output estimate for reasoning-heavy tasks.
When the beginning of a prompt is identical to a recent request, providers can serve it from a cache and bill it at a much lower rate, from 2.5% to 10% of the normal input price for the models in this calculator.
Claude Fable 5.2, Opus 5.5 and Sonnet 5.5 bill the full 1M-token context at the standard rate. Claude Haiku 5.5 has higher rates for prompts over 100,000 tokens, which the calculator applies automatically.
It varies widely. Every turn resends the system prompt and the conversation so far, so a short question late in a long chat can still send thousands of input tokens. Measure real prompts with the token counter.
The prices shown were checked in November 2026. Providers change prices and release models often, so confirm current rates on the official pricing pages linked above before you commit to a budget.