Free tool / cost
Which model is cheapest?
Every major model's price in one table, plus what your app would cost per month on each one.
No sign-up / prices checked 17 October 2026
Compare the bill before you pick.
What does your app do in a month?
Model calls, not users.
Prompt, history and retrieved text.
What the model writes back.
Only applies to models with a cache price.
Prices are US dollars per million tokens, checked 17 October 2026. Click a column to sort.
| Cached / 1M | |||||
|---|---|---|---|---|---|
| GPT-5 nanoOpenAI | $0.05 | $0.005 | $0.4 | 272K | $26.00$0.00026 per request |
| Llama 4 ScoutMeta · Priced at DeepInfra | $0.1 | — | $0.3 | 328K | $32.00$0.00032 per request · 1.2× the cheapest shown |
| Gemini 2.5 Flash-LiteGoogle | $0.1 | $0.01 | $0.4 | 1.05M | $36.00$0.00036 per request · 1.4× the cheapest shown |
| Claude Haiku 5.5Anthropic | $0.1 | $0.01 | $0.5 | 1M | $40.00$0.0004 per request · 1.5× the cheapest shown |
| GPT-4o miniOpenAI | $0.15 | $0.075 | $0.6 | 128K | $54.00$0.00054 per request · 2.1× the cheapest shown |
| Mistral SmallMistral | $0.15 | $0.015 | $0.6 | 262K | $54.00$0.00054 per request · 2.1× the cheapest shown |
| Llama 3.3 70BMeta · Priced at DeepInfra | $0.23 | — | $0.4 | 131K | $62.00$0.00062 per request · 2.4× the cheapest shown |
| Llama 4 MaverickMeta · Priced at DeepInfra | $0.2 | — | $0.8 | 1.05M | $72.00$0.00072 per request · 2.8× the cheapest shown |
| DeepSeek ChatDeepSeek | $0.28 | $0.028 | $0.42 | 131K | $72.80$0.00073 per request · 2.8× the cheapest shown |
| GPT-5.4 nanoOpenAI | $0.2 | $0.02 | $1.25 | 272K | $90.00$0.0009 per request · 3.5× the cheapest shown |
| DeepSeek V4 FlashDeepSeek | $0.3 | $0.006 | $1.20 | 1M | $108$0.00108 per request · 4.2× the cheapest shown |
| Gemini 3.1 Flash-LiteGoogle | $0.25 | $0.025 | $1.50 | 1.05M | $110$0.0011 per request · 4.2× the cheapest shown |
| GPT-5 miniOpenAI | $0.25 | $0.025 | $2 | 272K | $130$0.0013 per request · 5.0× the cheapest shown |
| GPT-4.1 miniOpenAI | $0.4 | $0.1 | $1.60 | 1.05M | $144$0.00144 per request · 5.5× the cheapest shown |
| Gemini 2.5 FlashGoogle | $0.3 | $0.03 | $2.50 | 1.05M | $160$0.0016 per request · 6.2× the cheapest shown |
| Mistral LargeMistral | $0.5 | $0.05 | $1.50 | 262K | $160$0.0016 per request · 6.2× the cheapest shown |
| Gemini 3.8 FlashGoogle | $0.75 | $0.075 | $3.75 | 1.05M | $300$0.003 per request · 12× the cheapest shown |
| GPT-5.4 miniOpenAI | $0.75 | $0.075 | $4.50 | 272K | $330$0.0033 per request · 13× the cheapest shown |
| Grok 4.3xAI | $1.25 | $0.2 | $2.50 | 1M | $350$0.0035 per request · 13× the cheapest shown |
| o4-miniOpenAI | $1.10 | $0.275 | $4.40 | 200K | $396$0.00396 per request · 15× the cheapest shown |
| Claude Haiku 4.5Anthropic | $1 | $0.1 | $5 | 200K | $400$0.004 per request · 15× the cheapest shown |
| DeepSeek V4 ProDeepSeek | $1.32 | $0.044 | $3.96 | 1M | $422$0.00422 per request · 16× the cheapest shown |
| Mistral MediumMistral | $1.50 | $0.15 | $7.50 | 262K | $600$0.006 per request · 23× the cheapest shown |
| Grok 4.5xAI | $2 | $0.3 | $6 | 500K | $640$0.0064 per request · 25× the cheapest shown |
| GPT-5OpenAI | $1.25 | $0.125 | $10 | 272K | $650$0.0065 per request · 25× the cheapest shown |
| Gemini 2.5 ProGoogle · Prompts over 200K tokens cost more | $1.25 | $0.125 | $10 | 1.05M | $650$0.0065 per request · 25× the cheapest shown |
| Gemini 3.5 FlashGoogle | $1.50 | $0.15 | $9 | 1.05M | $660$0.0066 per request · 25× the cheapest shown |
| GPT-4.1OpenAI | $2 | $0.5 | $8 | 1.05M | $720$0.0072 per request · 28× the cheapest shown |
| o3OpenAI | $2 | $0.5 | $8 | 200K | $720$0.0072 per request · 28× the cheapest shown |
| Claude Sonnet 5.5Anthropic | $2 | $0.1 | $10 | 1M | $800$0.008 per request · 31× the cheapest shown |
| GPT-4oOpenAI | $2.50 | $1.25 | $10 | 128K | $900$0.009 per request · 35× the cheapest shown |
| GPT-5.4OpenAI | $2.50 | $0.25 | $15 | 1.05M | $1,100$0.011 per request · 42× the cheapest shown |
| Claude Sonnet 4.6Anthropic | $3 | $0.3 | $15 | 1M | $1,200$0.012 per request · 46× the cheapest shown |
| GPT-5.6OpenAI | $4 | $0.4 | $20 | 922K | $1,600$0.016 per request · 62× the cheapest shown |
| Claude Opus 5.5Anthropic | $4 | $0.2 | $20 | 1M | $1,600$0.016 per request · 62× the cheapest shown |
| Claude Opus 4.7Anthropic | $5 | $0.5 | $25 | 1M | $2,000$0.02 per request · 77× the cheapest shown |
| GPT-5.5OpenAI | $5 | $0.5 | $30 | 1.05M | $2,200$0.022 per request · 85× the cheapest shown |
| Claude Fable 5.1Anthropic | $10 | $0.25 | $50 | 1M | $4,000$0.04 per request · 154× the cheapest shown |
Note Cheaper is not always better. A cheap model that fails or retries often can cost more per finished task. Test on your own prompts.
Prices are public list prices in US dollars per million tokens, taken from the LiteLLM community price list on 17 October 2026. They exclude batch, priority and regional pricing, tiered prices for very long prompts and free tiers. Llama models are open weights, so they are priced at one host (DeepInfra); other hosts charge differently. Confirm on the provider's page before you commit to a budget.
How it works
Three steps, no sign-up.
- 01
Describe your app
Requests per month, tokens in and out per request, and how much input a cache can serve.
- 02
Read the table
Each row shows list prices per million tokens and what your traffic would cost per month.
- 03
Sort and filter
Sort by any column, filter by provider, or search for a model.
Questions people ask
Where do the prices come from?
From the public LiteLLM model price list, checked on 17 October 2026. We show the date on the page. Providers change prices often, so always confirm on the provider's own page.
Why is monthly cost not just the price per million tokens?
Your bill is tokens in times the input price plus tokens out times the output price, for every request. This page does that maths with your numbers, and applies the cached-input price where the model has one.
Is the cheapest model the best choice?
Not always. A cheaper model that needs more retries, longer prompts or human fixes can cost more per finished task. Compare on your own prompts.
Why are some prices missing a cached rate?
Not every provider or model offers a cheaper price for cached input. A dash means there is no cache discount on record.
How do I see what I really spend?
Provider dashboards show totals. Tracify shows cost per run, per step and per model, so you can see which part of your agent is expensive.
More free tools
- Agent log checker
Paste a log or trace. See where it failed, what was slow and what wasted tokens.
- Token counter
Paste text. See the token count and price on GPT, Claude, Gemini and Llama.
- Prompt cost checker
Paste a prompt. Get a shorter, cheaper version and what it costs on each model.
- Context window checker
Does my document fit in each model's memory?
Now measure the real thing.
Send a first trace and see the true cost, speed and failures of every run.