Cost estimator
Turn a token count and a model choice into a monthly bill.
List prices in US dollars, standard tier. The estimate excludes tax, batch and volume discounts, and regional uplifts. The cache share applies only to a model that publishes a cached input price. A model that does not publish the dollar price this needs shows "Not published". A model the vendor lists as free is open weights you run on your own hardware, so its real cost is your servers, not an API bill.
| Model | Vendor | Per request | Per day | Per month |
|---|---|---|---|---|
| Qwen3.7-Flash (China platform) | Alibaba Cloud | < $0.0001 | $0.05 | $1.41 |
| Qwen3.7-Flash | Alibaba Cloud | < $0.0001 | $0.05 | $1.62 |
| Command R7B | Cohere | < $0.0001 | $0.06 | $1.91 |
| Ministral 3 (3B) | Mistral | < $0.0001 | $0.08 | $2.40 |
| Ministral 3 (8B) | Mistral | $0.0001 | $0.12 | $3.60 |
| Qwen3.8-Flash (China platform) | Alibaba Cloud | $0.0002 | $0.17 | $5.13 |
| Qwen3.8-Flash | Alibaba Cloud | $0.0002 | $0.22 | $6.48 |
| GLM-5.3-Flash | Zhipu AI | $0.0002 | $0.23 | $6.75 |
| Command R (08-2024) | Cohere | $0.0003 | $0.26 | $7.65 |
| Mistral Small 4 | Mistral | $0.0003 | $0.26 | $7.65 |
| Codestral | Mistral | $0.0004 | $0.42 | $12.60 |
| GPT-5.6 Luna | OpenAI | $0.0005 | $0.46 | $13.80 |
| Qwen3.7-Plus (China platform) | Alibaba Cloud | $0.0005 | $0.47 | $14.05 |
| GPT-5.4 nano | OpenAI | $0.0005 | $0.48 | $14.25 |
| MiniMax-M2.5 | MiniMax | $0.0005 | $0.51 | $15.30 |
| MiniMax-M2.7 | MiniMax | $0.0005 | $0.51 | $15.30 |
| MiniMax-M3 | MiniMax | $0.0005 | $0.51 | $15.30 |
| DeepSeek V4 Flash | DeepSeek | $0.0005 | $0.55 | $16.38 |
| DeepSeek V4 Flash Vision (experimental) | DeepSeek | $0.0005 | $0.55 | $16.38 |
| Qwen3.7-Plus | Alibaba Cloud | $0.0007 | $0.68 | $20.40 |
| Mistral Large 3 | Mistral | $0.0009 | $0.88 | $26.40 |
| Gemini 2.5 Flash | $0.0009 | $0.90 | $27.00 | |
| Gemini 3.5 Flash-Lite | $0.0009 | $0.90 | $27.00 | |
| GLM-4.7 | Zhipu AI | $0.0010 | $0.96 | $28.80 |
| MiniMax-M2.7 (high speed) | MiniMax | $0.0010 | $1.02 | $30.60 |
| Grok Build 0.1 | xAI | $0.0011 | $1.10 | $33.00 |
| Grok 4.3 | xAI | $0.0014 | $1.38 | $41.25 |
| GLM-5 | Zhipu AI | $0.0015 | $1.46 | $43.80 |
| Gemini 3.8 Flash | $0.0015 | $1.50 | $45.00 | |
| Kimi K2.6 | Moonshot AI | $0.0017 | $1.68 | $50.25 |
| Kimi K2.7 Code | Moonshot AI | $0.0017 | $1.68 | $50.25 |
| GPT-5.4 mini | OpenAI | $0.0017 | $1.72 | $51.75 |
| DeepSeek V4 Pro | DeepSeek | $0.0018 | $1.85 | $55.44 |
| Claude Haiku 4.5 | Anthropic | $0.0020 | $2.00 | $60.00 |
| GLM-5.1 | Zhipu AI | $0.0020 | $2.02 | $60.60 |
| GLM-5.2 | Zhipu AI | $0.0020 | $2.02 | $60.60 |
| GLM-5.3 | Zhipu AI | $0.0020 | $2.02 | $60.60 |
| Qwen3.7-Max (China platform) | Alibaba Cloud | $0.0023 | $2.31 | $69.31 |
| Qwen3.8-Max (China platform) | Alibaba Cloud | $0.0023 | $2.31 | $69.31 |
| Grok 4.5 | xAI | $0.0028 | $2.80 | $84.00 |
| Grok 4.6 | xAI | $0.0028 | $2.80 | $84.00 |
| Qwen3.8-Max | Alibaba Cloud | $0.0028 | $2.80 | $84.00 |
| Mistral Medium 3.5 | Mistral | $0.0030 | $3.00 | $90.00 |
| Kimi K2.7 Code (high speed) | Moonshot AI | $0.0034 | $3.35 | $100.50 |
| Gemini 3.5 Flash | $0.0034 | $3.45 | $103.50 | |
| Qwen3.7-Max | Alibaba Cloud | $0.0035 | $3.50 | $105.00 |
| Gemini 2.5 Pro | $0.0036 | $3.63 | $108.75 | |
| Claude Sonnet 5 | Anthropic | $0.0040 | $4.00 | $120.00 |
| Command R+ (08-2024) | Cohere | $0.0043 | $4.25 | $127.50 |
| Gemini 3.1 Pro Preview | $0.0046 | $4.60 | $138.00 | |
| GPT-5.6 Terra | OpenAI | $0.0046 | $4.60 | $138.00 |
| Kimi K3 | Moonshot AI | $0.0060 | $6.00 | $180.00 |
| GPT-5.6 Sol | OpenAI | $0.0080 | $8.00 | $240.00 |
| Claude Opus 5 | Anthropic | $0.01 | $10.00 | $300.00 |
| Claude Fable 5.1 | Anthropic | $0.02 | $20.00 | $600.00 |
| GPT-6 Astra | OpenAI | $0.02 | $20.00 | $600.00 |
| Command A+ | Cohere | Free (self-hosted) | Free (self-hosted) | Free (self-hosted) |
| GLM-4.7-FlashX (China platform) | Zhipu AI | Not published | Not published | Not published |
| GLM-5 (China platform) | Zhipu AI | Not published | Not published | Not published |
| GLM-5.1 (China platform) | Zhipu AI | Not published | Not published | Not published |
| GLM-5.2 (China platform) | Zhipu AI | Not published | Not published | Not published |
| GLM-5.3 (China platform) | Zhipu AI | Not published | Not published | Not published |
| GLM-5.3-Flash (China platform) | Zhipu AI | Not published | Not published | Not published |
| Kimi K2.6 (China platform) | Moonshot AI | Not published | Not published | Not published |
| Kimi K2.7 Code (China platform) | Moonshot AI | Not published | Not published | Not published |
| Kimi K2.7 Code (high speed) (China platform) | Moonshot AI | Not published | Not published | Not published |
| Kimi K3 (China platform) | Moonshot AI | Not published | Not published | Not published |
Prices checked 5 Sep 2026. See the full API pricing table and its sources.
This tool runs entirely in your browser. The prices are built into the page when the site is published, the arithmetic happens as you type, and nothing you enter is sent anywhere, stored or logged.
What this estimates
Give the tool the shape of your usage, how many requests you send a day, how long each one is, and how many days a month you run it, and it prints what each model would cost per request, per day and per month. The table holds every model this site prices, across every vendor, sorted so the cheapest monthly cost is at the top. Change a figure and the whole table re-sorts as you type.
How the monthly cost is calculated
The cost of one request is the input tokens times the model's input price per million tokens, plus the output tokens times its output price per million tokens, divided by one million. That per-request cost is multiplied by your requests per day, then by your days per month. If you set a cache share, it lowers the price of that fraction of the input, but only for a model that publishes a cached input rate: assuming a discount a vendor does not list would be inventing a number. The prices come straight from this site's API pricing page, where each one is stamped with the vendor page it was read from and the date it was checked. This page never types a price by hand.
Why some models show Not published
A few models publish a price only in another currency, and this tool never converts one currency into another: a made-up exchange rate would be a fabricated figure, which this site does not print. When the US dollar price the estimate needs is missing, the row shows "Not published", the same wording the price tables use, and sorts below every priced model rather than looking free.
What the estimate leaves out
These are list prices on the standard tier. The estimate excludes tax, batch and volume discounts, and the regional uplifts some vendors add, a few of which a model's own notes on the pricing page spell out. It is a way to compare models on the same workload, not a quote for your bill.
Questions people ask
Where do the prices come from?
From this site's own API pricing page. Every figure there carries the vendor page it was read from and the date it was checked, and the same figures are built into this tool when the site is published. The tool reads only the US dollar prices: a model that prices in another currency is shown as not published rather than converted at a rate we would have to invent.
How is the monthly cost worked out?
The cost of one request is the input tokens times the input price per million tokens, plus the output tokens times the output price per million tokens, divided by one million. That is multiplied by the requests per day and then by the days per month. The cache share lowers the price of the cached part of the input, but only for a model that publishes a cached input rate.
Why do some rows say Not published?
Because that model does not publish, in US dollars, the price the estimate needs. Some models are priced only in another currency, and this tool never converts one price into another. A row like that shows Not published, the same wording the price tables use, and sorts to the bottom rather than appearing to be free.
Is this my actual bill?
No, it is an estimate from list prices. It uses the standard tier and leaves out tax, batch and volume discounts, and regional uplifts, some of which a model's own notes on the pricing page describe. Your real usage, and any discount you have negotiated, will move the number.
Does this page send or store what I enter?
No. The prices are built into the page and the arithmetic runs in your browser. The page makes no request as you type, keeps no copy of your figures after you close the tab, and loads no analytics.