StackLift
Updated Nov 22, 2026, 3:07 PM UTC

LLM API pricing, updated every hour.

What Claude, GPT, Gemini, Llama, Mistral, DeepSeek and Qwen cost per million tokens today, and what your usage would cost each month.

At a glance

The flagship model from each provider

The models most products start with, cheapest first. Blended assumes three input tokens for every output token.

ProviderModelInput / 1MOutput / 1MBlended / 1M
Gemma (Google)Gemma 4 26B A4B $0.068$0.225$0.107
Qwen / DeepSeekQwen3.8 27B$0.035$1.50$0.401
Llama and open modelsLlama 4 Maverick$0.35$1.00$0.513
MistralMistral Large 4$0.70$2.10$1.05
Claude (Anthropic)Claude Sonnet 5.5$2.00$10.00$4.00
AWS BedrockClaude Sonnet 5.5$2.00$10.00$4.00
Gemini (Google)Gemini 4 Neon$3.00$15.00$6.00
Google Vertex AIGemini 4 Neon$3.00$15.00$6.00
Not chosen yetAverage of Claude, GPT, and Gemini$3.33$18.33$7.08
OpenAI GPT (ChatGPT)GPT-5.5$5.00$30.00$11.25
Azure OpenAIGPT-5.5$5.00$30.00$11.25
Fine-tuned hosted modelGPT-5.5$7.50$45.00$16.88

Every model, and your monthly cost

Enter your monthly token use and sort by “Your month” to find the cheapest model for it. 272 models from 7 providers, as of Nov 22, 2026, 3:07 PM UTC.

About 10M in / 3M out is a support assistant handling ~6,000 conversations a month.

272 models
Llama 3.1 8B InstructMeta (Llama) · meta-llama/llama-3.1-8b-instruct$0.02$0.04131K$0.32
Mistral NemoMistral · mistralai/mistral-nemo$0.029$0.03131K$0.38
Gemma 3 4BGoogle (Gemini, Gemma) · google/gemma-3-4b-it$0.04$0.05131K$0.55
gpt-oss-20b (batch)OpenAI (GPT, ChatGPT) · openai/gpt-oss-20b:batch$0.024$0.112131K$0.58
Qwen3.7 FlashQwen (Alibaba) · qwen/qwen3.7-flash$0.03$0.131M$0.69
gpt-oss-120b (batch)OpenAI (GPT, ChatGPT) · openai/gpt-oss-120b:batch$0.03$0.136131K$0.70
Mistral Small 3Mistral · mistralai/mistral-small-24b-instruct-2501$0.05$0.0833K$0.74
DeepSeek V4.1 Flash 1017 (batch)DeepSeek · deepseek/deepseek-v4.1-flash-1017:batch$0.035$0.1491M$0.80
GPT-5 Nano (batch)OpenAI (GPT, ChatGPT) · openai/gpt-5-nano:batch$0.025$0.20400K$0.85
DeepSeek V4 Flash 0731 (batch)DeepSeek · deepseek/deepseek-v4-flash-0731:batch$0.048$0.1441M$0.91
DeepSeek V4.1 Flash (batch)DeepSeek · deepseek/deepseek-v4.1-flash:batch$0.05$0.151M$0.95
Gemma 3 12BGoogle (Gemini, Gemma) · google/gemma-3-12b-it$0.05$0.15131K$0.95
DeepSeek V4.1 FlashDeepSeek · deepseek/deepseek-v4.1-flash$0.05$0.1991M$1.10
GPT-4.1 Nano (batch)OpenAI (GPT, ChatGPT) · openai/gpt-4.1-nano:batch$0.05$0.201M$1.10
Claude Haiku 5.5 (batch)Anthropic (Claude) · anthropic/claude-haiku-5.5:batch$0.05$0.251M$1.25
GPT-6 Luna Pro (batch)OpenAI (GPT, ChatGPT) · openai/gpt-6-luna-pro:batch$0.05$0.251.1M$1.25
GPT-6 Luna (batch)OpenAI (GPT, ChatGPT) · openai/gpt-6-luna:batch$0.05$0.251.1M$1.25
Ministral 3 3B 2512Mistral · mistralai/ministral-3b-2512$0.10$0.10131K$1.30
DeepSeek V4 Flash 0731DeepSeek · deepseek/deepseek-v4-flash-0731$0.044$0.301M$1.34
Gemma 4 26B A4B Google (Gemini, Gemma) · google/gemma-4-26b-a4b-it$0.068$0.225262K$1.35
Mistral Small 3.2 24BMistral · mistralai/mistral-small-3.2-24b-instruct$0.075$0.20256K$1.35
Qwen3.5-FlashQwen (Alibaba) · qwen/qwen3.5-flash-02-23$0.065$0.261M$1.43
DeepSeek V4 Flash 0423DeepSeek · deepseek/deepseek-v4-flash$0.09$0.181M$1.44
Llama 3.2 3B InstructMeta (Llama) · meta-llama/llama-3.2-3b-instruct$0.05$0.33131K$1.49
Qwen3 Coder 30B A3B InstructQwen (Alibaba) · qwen/qwen3-coder-30b-a3b-instruct$0.07$0.28262K$1.54
1–25 of 272Page 1 of 11

List prices in USD from the OpenRouter model catalog; your provider's own prices, discounts and taxes may differ. Context is the most tokens a model reads at once (128K ≈ a 300-page book).

Plan your AI product

Know the full running cost, not just tokens

A real AI feature also needs hosting, a vector database, monitoring and the build itself. The free estimate adds them up for your project in about five minutes.

Questions about LLM pricing

How often are these LLM prices updated?

Every hour. StackLift checks the public OpenRouter model catalog every hour and shows the list price per million tokens for each model. The last update was Nov 22, 2026, 3:07 PM UTC.

What does "per million tokens" mean?

Providers charge for the text a model reads (input tokens) and writes (output tokens). A token is about three-quarters of an English word, so a million tokens is roughly 750,000 words. Output usually costs several times more than input.

How much does an AI chatbot cost to run each month?

For a support assistant handling 10,000 conversations a month (about 1,500 tokens in and 500 out each, so 15M input and 5M output tokens), the model costs about $2.14 a month with Gemma 4 26B A4B , the lowest-priced of the flagship models in this table. Hosting, a vector database and monitoring are extra; the StackLift estimate adds them for your project.

Which LLM should I use for my product?

Start from the task, not the price: reasoning-heavy work suits the larger models, while classification, extraction and routine replies run well on smaller, cheaper ones. Many products mix both. The free StackLift estimate recommends models for your use case and shows the monthly running cost.

Are fine-tuning, hosting or self-hosted models included?

No. These are pay-as-you-go API list prices. Self-hosting an open model (Llama, Qwen, DeepSeek) trades token costs for GPU servers; batch and cached-input discounts can lower API costs further.

The complete guide

How LLM API pricing works

Updated November 18, 2026 · Published by , Founder, StackLift AI. Written with AI assistance; how we make content.

Short answer

LLM APIs charge per token: one rate for the text you send in, a higher rate for the text the model writes back, and sometimes extra for reasoning, with discounts for cached input and batch jobs. Your monthly bill is tokens per request times requests per month times those rates. The live prices are in the table above.

How does LLM API pricing work?

Providers meter usage in tokens, small pieces of text. Prices are quoted per million tokens, with one rate for input and another for output. The prices shown above on this page are list prices that refresh through the day, so this guide explains how to use them rather than repeat them.

The parts of an LLM bill
PartWhat it isWhat to watch
Input tokensEverything you send: instructions, history, documents, the question.Long prompts and pasted documents add up on every call.
Output tokensWhat the model writes back.Usually priced higher than input, so long answers cost more.
Context windowThe most text a model can read in one request.A bigger window lets you send more, but you pay for what you send.
Cached inputRepeated prompt content the provider has stored.Priced lower than fresh input when the same prefix repeats.
Batch processingJobs submitted for later, not instant, results.Often discounted, and only fits work that can wait.
Reasoning tokensHidden thinking some models do before answering.Billed like output, and easy to miss in an estimate.

Rules differ by provider and change over time. Always confirm discounts and limits on the provider's own pricing page before you commit to a budget.

How do you estimate a monthly LLM bill?

Work from your real workload, not from a guess at tokens. A rough estimate built on honest inputs beats a precise one built on wishes.

  1. Pick one task, such as answering a support question or summarizing a document.
  2. Take a sample of real requests and count the input tokens per request, including instructions and any retrieved text.
  3. Count the output tokens of typical answers, and add room for reasoning if the model thinks before replying.
  4. Multiply by the number of requests per month, with a low, expected and high case.
  5. Apply input and output rates from the table above for each candidate model.
  6. Reduce the input share that repeats (cached) and the share that can run as a batch.
  7. Add retries, failed calls and test traffic, then keep a margin for growth.

The explorer above does the arithmetic when you enter your monthly token use. If you are budgeting an agent, which makes several calls per task, see the AI agent costs page, and for a full build plus running cost use the free estimate.

How do you choose a model for a workload?

Choose by the task, then check the price. Test two or three models on your own examples and compare quality, speed and cost per finished task. A model that costs more per token but needs fewer retries can cost less overall.

Matching workload to model type
WorkloadWhat matters mostModel type to try first
Classification, tagging, routingSpeed and low cost at volumeA small, fast model
Extraction from documentsAccuracy on your formatsA small or mid-size model, tested on real files
Customer support repliesTone, grounding in your content, safetyA mid-size model with retrieval
Multi-step agentsReliable tool use and planningA larger model, with cheaper ones for simple steps
Code, analysis, long reasoningReasoning qualityA larger or reasoning model
Long documentsContext window and recallA model with a large window, or retrieval to shorten the input
Sensitive dataWhere the data is processedA model or hosting option that fits your data rules

Many products use more than one model: a small one for routine steps and a larger one for the hard ones. That pattern is called model routing. Provider-specific pages with live prices are linked from the explorer above for the major models.

What are the ways to pay less for LLM APIs?

Cost cuts that keep quality come from sending less, reusing more and picking the smaller model where it is enough.

Levers that lower an LLM bill
LeverHow it savesTrade-off
Shorter promptsFewer input tokens on every call.Needs careful editing and re-testing.
Prompt cachingRepeated prefixes cost less.Only helps when the start of the prompt stays the same.
Batch processingDiscounted rates for jobs that can wait.Results are not instant.
Model routingEasy tasks go to cheaper models.Needs a router and evaluation of each path.
Retrieval instead of long contextSend only the relevant passages.Needs a good retrieval setup.
Output limitsCaps long or runaway answers.Can cut answers short if set too low.
Response reuseStore answers to repeated questions.Needs rules for when an answer goes stale.

The LLM cost optimization playbook puts these in order, and the glossary entry on AI cost optimization defines them.

What does an LLM price list not include?

List prices cover model usage only. A working product has other costs, and they often outweigh the tokens at low volume.

  • Build work: prompts, retrieval, evaluation and integration with your systems.
  • Hosting, storage and a vector database if you use retrieval.
  • Monitoring, logging and the tools to review quality over time.
  • Guardrails, security review and handling of sensitive data.
  • Self-hosting an open model: you swap token fees for GPU servers and the people to run them.
  • Fine-tuning or training, which are priced separately.

For the build side, see LLM integration and its cost guide. If your data cannot leave your own environment, read about private LLM development.

How should you read the price tables on this page?

Prices are list prices in US dollars from a public source, so your provider's own prices, discounts and taxes may differ. Use the tables to compare models against each other and to set a budget range. Do not treat a single price as a promise.

A cheaper token rate does not mean a cheaper product. Compare the cost to finish a task at your quality bar, and re-check when providers change prices or release new models. A short monthly review of usage against budget catches drift early.

StackLift tracks model prices so the free estimate can recommend models for your use case and show the monthly running cost next to the build cost. Every estimate is reviewed by a senior architect before it becomes a commitment.

Questions people ask

What is a token in an LLM API?

A token is a small piece of text, often part of a word. Providers count tokens for what you send and what the model returns, and charge per million tokens. Read more in the tokens glossary entry.

Why is output more expensive than input?

Generating text takes more computation than reading it, so most providers price output tokens higher than input tokens. The exact gap differs by model, so check the table above.

What are reasoning tokens?

Some models think before they answer. That hidden thinking is usually billed like output even though you do not see it, so allow for it when you estimate a bill.

What does prompt caching do?

It stores a repeated start of your prompt, such as fixed instructions, so later calls read it at a lower rate. See prompt caching.

Is a bigger context window worth paying for?

Only if you need it. You pay for every token you send, so a big window used carelessly raises cost. Retrieval often reaches the same goal with less input. Read about the context window.

Which model should I use?

Test two or three on your own examples and compare cost per finished task, not price per token. Many products use a small model for routine steps and a larger one for hard ones.

Can I use these prices as a quote?

No. They are list prices for planning. Your provider's pricing page and your contract set the real rates.