What an AI job costs
The price of each text model from 7 makers, for 1 million tokens: input, cached input, output, batch and long prompts. Sort them, put up to 4 side by side, then turn a price into the bill for a real job.
Prices read from each maker's page on 14 October 2026
The price list
US dollars for 1 million tokens, from each maker's own price page. A million tokens is about 750,000 English words. Tick up to 4 models to put them side by side.
16 of 89 models
| Compare | More | ||||||
|---|---|---|---|---|---|---|---|
| GPT-6 AstraOpenAI | $10 | $1 | $50 | $20 | 1.05M | ||
| GPT-6.1 SolOpenAI | $2 | $0.10 | $10 | $4 | 1.05M | ||
| GPT-6 LunaOpenAI | $0.10 | $0.01 | $0.50 | $0.20 | 1.05M | ||
| GPT-6 SolOpenAI | $2 | $0.20 | $10 | $4 | 1.05M | ||
| GPT-5.6 SolOpenAI | $4 | $0.40 | $20 | $8 | 1.05M | ||
| GPT-5.6 TerraOpenAI | $2 | $0.20 | $12 | $4.50 | 1.05M | ||
| GPT-5.6 LunaOpenAI | $0.20 | $0.02 | $1.20 | $0.45 | 1.05M | ||
| gpt-5.5OpenAI | $5 | $0.50 | $30 | $11.30 | 1.05M | ||
| gpt-5.5-proOpenAI | $30 | none | $180 | $67.50 | 1.05M | ||
| gpt-5.4OpenAI | $2.50 | $0.25 | $15 | $5.63 | 1.05M | ||
| gpt-5.4-miniOpenAI | $0.75 | $0.075 | $4.50 | $1.69 | 400K | ||
| gpt-5.4-nanoOpenAI | $0.20 | $0.02 | $1.25 | $0.463 | 400K | ||
| gpt-5.4-proOpenAI | $30 | none | $180 | $67.50 | 1.05M | ||
| gpt-5.2OpenAI | $1.75 | $0.175 | $14 | $4.81 | 400K | ||
| gpt-5.2-proOpenAI | $21 | none | $168 | $57.80 | 400K | ||
| gpt-5.1OpenAI | $1.25 | $0.125 | $10 | $3.44 | 400K | ||
| gpt-5OpenAI | $1.25 | $0.125 | $10 | $3.44 | 400K | ||
| gpt-5-miniOpenAI | $0.25 | $0.025 | $2 | $0.688 | 400K | ||
| gpt-5-nanoOpenAI | $0.05 | $0.005 | $0.40 | $0.138 | 400K | ||
| gpt-5-proOpenAI | $15 | none | $120 | $41.30 | 400K | ||
| gpt-4.1OpenAI | $2 | $0.50 | $8 | $3.50 | 1.05M | ||
| gpt-4.1-miniOpenAI | $0.40 | $0.10 | $1.60 | $0.70 | 1.05M | ||
| gpt-4.1-nanoOpenAI | $0.10 | $0.025 | $0.40 | $0.175 | 1.05M | ||
| gpt-4oOpenAI | $2.50 | $1.25 | $10 | $4.38 | 128K | ||
| gpt-4o-2024-05-13OpenAI | $5 | none | $15 | $7.50 | n/a | ||
| gpt-4o-miniOpenAI | $0.15 | $0.075 | $0.60 | $0.262 | 128K | ||
| o1OpenAI | $15 | $7.50 | $60 | $26.30 | 200K | ||
| o1-proOpenAI | $150 | none | $600 | $263 | 200K | ||
| o3-proOpenAI | $20 | none | $80 | $35 | 200K | ||
| o3OpenAI | $2 | $0.50 | $8 | $3.50 | 200K | ||
| o4-miniOpenAI | $1.10 | $0.275 | $4.40 | $1.93 | 200K | ||
| o3-miniOpenAI | $1.10 | $0.55 | $4.40 | $1.93 | 200K | ||
| gpt-4-turbo-2024-04-09OpenAI | $10 | none | $30 | $15 | n/a | ||
| gpt-4-0613OpenAI | $30 | none | $60 | $37.50 | n/a | ||
| gpt-3.5-turboOpenAI | $0.50 | none | $1.50 | $0.75 | 16K | ||
| gpt-3.5-turbo-0125OpenAI | $0.50 | none | $1.50 | $0.75 | n/a | ||
| gpt-3.5-turbo-1106OpenAI | $1 | none | $2 | $1.25 | n/a | ||
| gpt-3.5-turbo-instructOpenAI | $1.50 | none | $2 | $1.63 | 4K | ||
| davinci-002OpenAI | $2 | none | $2 | $2 | n/a | ||
| babbage-002OpenAI | $0.40 | none | $0.40 | $0.40 | n/a | ||
| Claude Fable 5.1Anthropic | $10 | $0.25 | $50 | $20 | 1M | ||
| Claude Mythos 5.1Anthropic | $10 | $0.25 | $50 | $20 | 1M | ||
| Claude Fable 5Anthropic | $10 | $1 | $50 | $20 | 1M | ||
| Claude Mythos 5Anthropic | $10 | $1 | $50 | $20 | 1M | ||
| Claude Opus 5.5Anthropic | $4 | $0.20 | $20 | $8 | 1M | ||
| Claude Opus 5Anthropic | $5 | $0.50 | $25 | $10 | 1M | ||
| Claude Opus 4.8Anthropic | $5 | $0.50 | $25 | $10 | 1M | ||
| Claude Opus 4.7Anthropic | $5 | $0.50 | $25 | $10 | 1M | ||
| Claude Opus 4.6Anthropic | $5 | $0.50 | $25 | $10 | 1M | ||
| Claude Opus 4.5Anthropic | $5 | $0.50 | $25 | $10 | 200K | ||
| Claude Opus 4.1Anthropic · retired | $15 | $1.50 | $75 | $30 | n/a | ||
| Claude Opus 4Anthropic · retired | $15 | $1.50 | $75 | $30 | n/a | ||
| Claude Sonnet 5.5Anthropic | $2 | $0.10 | $10 | $4 | 1M | ||
| Claude Sonnet 5Anthropic | $2 | $0.20 | $10 | $4 | 1M | ||
| Claude Sonnet 4.6Anthropic | $3 | $0.30 | $15 | $6 | 1M | ||
| Claude Sonnet 4.5Anthropic | $3 | $0.30 | $15 | $6 | 200K | ||
| Claude Sonnet 4Anthropic · retired | $3 | $0.30 | $15 | $6 | n/a | ||
| Claude Haiku 5.5Anthropic | $0.10 | $0.01 | $0.50 | $0.20 | 1M | ||
| Claude Haiku 4.5Anthropic | $1 | $0.10 | $5 | $2 | 200K | ||
| Claude Haiku 3.5Anthropic · retired | $0.80 | $0.08 | $4 | $1.60 | n/a | ||
| Gemini 4 ArgonGoogle | $2 | $0.10 | $10 | $4 | n/a | ||
| Gemini 3.8 FlashGoogle | $0.75 | $0.075 | $3.75 | $1.50 | 1.05M | ||
| Gemini 3.6 FlashGoogle | $0.75 | $0.075 | $3.75 | $1.50 | 1.05M | ||
| Gemini 3.5 Flash-LiteGoogle | $0.30 | $0.03 | $2.50 | $0.85 | 1.05M | ||
| Gemini 3.1 Pro PreviewGoogle | $2 | $0.20 | $12 | $4.50 | 1.05M | ||
| Gemini 3.1 Flash-LiteGoogle | $0.25 | $0.025 | $1.50 | $0.563 | 1.05M | ||
| Gemini 3 Flash PreviewGoogle | $0.50 | $0.05 | $3 | $1.13 | 1.05M | ||
| Gemini 2.5 ProGoogle | $1.25 | $0.125 | $10 | $3.44 | 1.05M | ||
| Gemini 2.5 FlashGoogle | $0.30 | $0.03 | $2.50 | $0.85 | 1.05M | ||
| Gemini 2.5 Flash-LiteGoogle | $0.10 | $0.01 | $0.40 | $0.175 | 1.05M | ||
| Grok 4.7xAI | $2 | $0.50 | $6 | $3 | 500K | ||
| Grok 4.6xAI | $2 | $0.50 | $6 | $3 | 500K | ||
| Grok 4.5xAI | $2 | $0.30 | $6 | $3 | 500K | ||
| Grok 4.3xAI | $1.25 | $0.20 | $2.50 | $1.56 | 1M | ||
| grok-4.20-0309-reasoningxAI | $1.25 | $0.20 | $2.50 | $1.56 | 1M | ||
| grok-4.20-0309-non-reasoningxAI | $1.25 | $0.20 | $2.50 | $1.56 | 1M | ||
| grok-build-0.1xAI | $1 | $0.20 | $2 | $1.25 | 256K | ||
| grok-4.20-multi-agent-0309xAI | $1.25 | $0.20 | $2.50 | $1.56 | 1M | ||
| DeepSeek V4 ProDeepSeek | $1.32 | $0.044 | $3.96 | $1.98 | 1M | ||
| DeepSeek V4.1 FlashDeepSeek | $0.30 | $0.006 | $1.20 | $0.525 | 1M | ||
| Kimi K3Moonshot AI | $3 | $0.30 | $15 | $6 | 1.05M | ||
| Kimi K2.7 CodeMoonshot AI | $0.95 | $0.19 | $4 | $1.71 | 262K | ||
| Kimi K2.7 Code HighspeedMoonshot AI | $1.90 | $0.38 | $8 | $3.42 | 262K | ||
| Kimi K2.6Moonshot AI | $0.95 | $0.16 | $4 | $1.71 | 262K | ||
| Qwen3.8-MaxAlibaba | $2 | $0.25 | $6 | $3 | n/a | ||
| Qwen3.8-FlashAlibaba | $0.15 | $0.016 | $0.47 | $0.23 | n/a | ||
| Qwen3.7-MaxAlibaba | $2.50 | $0.50 | $7.50 | $3.75 | n/a | ||
| Qwen3.7-PlusAlibaba | $0.32 | $0.064 | $1.28 | $0.56 | n/a | ||
| Qwen3.7-FlashAlibaba | $0.03 | $0.006 | $0.13 | $0.055 | n/a |
Side by side
Tick 2 to 4 models in the list above, or start from one of these sets.
Put a job on the meter
A price per token is half of a bill. The other half is how many tokens the job takes. Pick a job or type your own numbers, and see what it costs on each model.
The job, for 1 run
A short question and a few paragraphs back.
Text the model was sent a moment ago, such as earlier turns of a long task, is charged at the cached price.
The bill, cheapest first · the newest models in the list
- Input
- Cached input
- Output
- Thinking
Qwen3.8-FlashAlibaba
0.028 cents
input 0.0045 cents · output 0.024 cents84 cents a month
GPT-6 LunaOpenAI
0.028 cents
input 0.003 cents · output 0.025 cents84 cents a month
Claude Haiku 5.5Anthropic
0.028 cents
input 0.003 cents · output 0.025 cents84 cents a month
DeepSeek V4.1 FlashDeepSeek
0.069 cents
input 0.009 cents · output 0.06 cents$2.07 a month
Gemini 3.5 Flash-LiteGoogle
0.13 cents
input 0.009 cents · output 0.13 cents$4.02 a month
Gemini 3.8 FlashGoogle
0.21 cents
input 0.022 cents · output 0.19 cents$6.30 a month
DeepSeek V4 ProDeepSeek
0.24 cents
input 0.04 cents · output 0.2 cents$7.13 a month
Grok 4.7xAI
0.36 cents
input 0.06 cents · output 0.3 cents$10.80 a month
Qwen3.8-MaxAlibaba
0.36 cents
input 0.06 cents · output 0.3 cents$10.80 a month
GPT-6.1 SolOpenAI
0.56 cents
input 0.06 cents · output 0.5 cents$16.80 a month
Claude Sonnet 5.5Anthropic
0.56 cents
input 0.06 cents · output 0.5 cents$16.80 a month
Gemini 4 ArgonGoogle
0.56 cents
input 0.06 cents · output 0.5 cents$16.80 a month
Kimi K3Moonshot AI
0.84 cents
input 0.09 cents · output 0.75 cents$25.20 a month
Claude Opus 5.5Anthropic
1.1 cents
input 0.12 cents · output 1 cents$33.60 a month
GPT-6 AstraOpenAI
2.8 cents
input 0.3 cents · output 2.5 cents$84.00 a month
Claude Fable 5.1Anthropic
2.8 cents
input 0.3 cents · output 2.5 cents$84.00 a month
List prices with no discounts. Where a maker publishes no cached price, cached input is charged as input. The bars share one scale.
Recent price changes
A price can change after a model launches. Open a model in the list to see every change recorded for it.
7 October 2026
Claude Sonnet 5.5
Anthropic
Cached input went from $0.20 to $0.10 per million tokens. The price of a cache write did not change.
Read it in What changed21 August 2026
GPT-5.6 Sol
OpenAI
The price became $4 per million input tokens and $20 per million output tokens, which OpenAI gives as 20% lower for input and 33% lower for output. OpenAI calls it promotional pricing, available at least until 21 November 2026.
Read it in What changed30 July 2026
GPT-5.6 Terra
OpenAI
The price fell by 20%.
Read it in What changed30 July 2026
GPT-5.6 Luna
OpenAI
The price fell by 80%.
Read it in What changed
4 things that decide the bill
The price is half of it
The other half is how many tokens the job takes. 2 models with the same price per token can hand you different bills for the same job.
Writing costs more than reading
On the newest models in the list, a token the model writes costs between 3 and 8 times as much as a token it reads.
Thinking is billed as writing
When a model thinks before it answers, those tokens are charged at the output price, even though you never read them.
A long task reads the same text again and again
An agent sends the conversation so far on every turn. That is how a task that never grows past 120,000 tokens can process 2.8 million.
What each column means
- Input
- What you pay for the tokens the model reads: your prompt, the files you attach and the conversation so far.
- Cached input
- What you pay when the model reads text it was sent a moment ago. It is a fraction of the input price, which is why long tasks cost less than the input price suggests.
- Cache write
- Some makers charge more than the input price the first time a piece of text is stored for reuse. It is shown where a maker prints it.
- Output
- What you pay for the tokens the model writes, the answer and any thinking before it.
- Blend 3:1
- Our own figure for ranking: the price of 1 million tokens when 3 are read for each 1 written. No maker charges it.
- Batch
- A lower price for work that can wait: you send the requests together and collect the answers later. On this list it is 50% to 100% of the standard price.
- Long prompts
- Some models charge more once a prompt passes a size, such as 200,000 tokens. Each model’s details give the size.
- Window
- The most tokens a model can hold at once: your prompt, the conversation so far and its answer.
Where the prices come from
Every figure is copied from the maker's own price page, never from a list someone else keeps. Prices change often. If a figure here and the maker's page disagree, the maker's page is right.
| Maker | Models listed | Page read on | The page |
|---|---|---|---|
| OpenAI | 40 | 14 October 2026 | developers.openai.com |
| Anthropic | 20 | 14 October 2026 | platform.claude.com |
| 10 | 14 October 2026 | ai.google.dev | |
| xAI | 8 | 14 October 2026 | docs.x.ai |
| DeepSeek | 2 | 14 October 2026 | api-docs.deepseek.com |
| Moonshot AI | 4 | 14 October 2026 | platform.kimi.ai |
| Alibaba | 5 | 14 October 2026 | qwencloud.com |
What the calculator leaves out
- The first write to the cache. The calculator charges cached text at the cached price. Storing it the first time costs more on some models. Caching is still what makes long tasks affordable: reading the 40-turn agent task on Claude Opus 5.5 costs $11.20 with no cache and $1.62 with 90% of it cached, the figure in Anthropic's worked example.
- Tools, pictures and sound. Web search, running code, images and audio have price lists of their own.
- Discounts. Off-peak hours, promotions and prices agreed with a maker. A model's details say where its price is dated or depends on the hour.
- The chat apps. ChatGPT, Claude and Gemini charge by plan, not by token. Tokens still decide when you hit a limit.