AI task costs · in pounds
How much AI costs to run ten everyday business tasks
For the ten jobs below, the AI model bill runs from under 1p to £561.72 a month, and 8 of the 10 cost under £10 a month in the middle of the range. The dearest is a support chatbot answering 10,000 chats: £5.39 a month on the cheapest of 24 models, £561.72 on the dearest. VAT not included.
These tables compare price only, not quality, and cover the model alone. What a first AI build costs a UK business is in our AI automation cost guide.
Prices of from OpenRouter's public API price list · converted at the ECB reference rate of ($1 = £0.7454) · Urbano DX
Ten tasks: the cheapest, middle and dearest model
| Task | Volume | Cheapest a month | Middle of the range | Dearest a month |
|---|---|---|---|---|
| Summarise a 50-page PDF | 20 PDFs a month | 3.4pDeepSeek V4.1 Flash | 39pMistral Large 4 | £5.81GPT-6 Astra |
| Draft replies to 1,000 customer emails | 1,000 emails a month | 22pDeepSeek V4.1 Flash | £2.01Mistral Large 4 | £32.42GPT-6 Astra |
| Run a support chatbot: 10,000 chats a month | 10,000 chats a month | £5.39DeepSeek V4.1 Flash | £32.50Llama 3.3 70B Instruct | £561.72GPT-6 Astra |
| Classify 10,000 support tickets | 10,000 tickets a month | 33pDeepSeek V4.1 Flash | £3.70Mistral Large 4 | £55.90GPT-6 Astra |
| Extract fields from 1,000 scanned invoices | 1,000 invoices a month | 13pDeepSeek V4.1 Flash | £1.57Gemini 3.8 Flash | £19.01GPT-6 Astra |
| Transcribe 100 hours of calls | 100 hours a month | 89pNemotron 3.5 ASR Streaming Multilingual 0.6B | £13.42Voxtral Mini Transcribe | £71.56Chirp 3 |
| Write 500 product descriptions | 500 descriptions a month | 6.1pDeepSeek V4.1 Flash | 42pMistral Large 4 | £7.45GPT-6 Astra |
| Translate a 20,000-word website | 2,000 changed words a month | 0.15pClaude Haiku 5.5 | 0.68pGemini 3.5 Flash Lite | 15pGPT-6 Astra |
| Generate 100 marketing images | 100 images a month | 52pRecraft V4.1 Flash | £2.61Seedream 5.0 Lite | £15.65Recraft V4.1 Pro |
| Answer questions over 10,000 internal documents | 2,000 questions a month | 58pDeepSeek V4.1 Flash | £5.48Mistral Large 4 | £87.21GPT-6 Astra |
The middle of the range is the median model of each task's list. Open a task for every model, the cost of a single job and the workload behind it.
Questions about AI running costs
How much does an AI chatbot cost per month?
In model charges, a website support chatbot handling 10,000 chats a month costs £5.62 with Claude Haiku 5.5 or GPT-6 Luna, £10.64 with gpt-oss-120b, £107.16 with Claude Sonnet 5.5 and £561.72 with GPT-6 Astra. That assumes four customer messages per chat and a prompt cache. A chatbot platform, the build and staff time are extra.
How much does AI transcription cost per hour?
Through a speech-to-text API, an hour of recorded calls costs from 0.89p with OpenAI's Whisper Large V3 Turbo to 72p with Google's Chirp 3. OpenAI's GPT Transcribe costs 20p an hour and Deepgram's Nova-3 19p. At 100 hours a month that is 89p to £71.56. Speaker labels and word timings can cost extra with some providers.
What does the ChatGPT, Claude or Gemini API cost for a business task?
OpenAI, Anthropic and Google each sell cheap and dear models, so the model matters more than the company. Summarising 20 fifty-page PDFs a month costs 5.8p with GPT-6 Luna or Claude Haiku 5.5, 48p with Gemini 3.8 Flash, £1.16 with GPT-6.1 Sol, Claude Sonnet 5.5 or Gemini 4 Argon, and £5.81 with GPT-6 Astra.
How much do OpenAI embeddings cost for searching company documents?
Embedding 10,000 documents of about 1,100 words (15 million tokens) costs 22p once with OpenAI's Text Embedding 3 Small and £1.45 with Text Embedding 3 Large; updates and question embeddings then cost 1.3p and 8.2p a month. Answering 2,000 staff questions a month adds 58p to £87.21, depending on the model. Vector storage and search are extra.
How much does AI image generation cost?
Through an image API, one image of about one megapixel costs 0.52p with Recraft V4.1 Flash, 2.3p with FLUX.2 Pro and 16p with Recraft V4.1 Pro, so 100 images a month cost 52p to £15.65, rejected attempts included. OpenAI's GPT Image and Google's Gemini image models bill in tokens, so they are left out.
How the numbers are worked out
- Each workload is written out on its page: tokens in and out, the share read from a prompt cache, audio minutes or images. Cost = workload × the model's listed rate, so you can check any figure with a calculator.
- Rates are OpenRouter's API prices in our snapshot of , the same ones on our model pricing pages. Buying from the developer directly can cost a little more or less. When rates move, the price change log records it.
- US dollars become pounds at the European Central Bank reference rate of : €1 = $1.1193 = £0.8343, so $1 = £0.7454. VAT, card fees and top-up fees are not included.
- About 0.75 English words make one token. Each model's tokenizer counts the same text a little differently, so treat a figure as a close estimate, not a quote.
- Models that reason before answering can add thinking tokens, billed as output. These figures count none, so a model with thinking switched on can cost more.
- This compares price only. It says nothing about which model does the job well: a cheap model that gets a tenth of the answers wrong can cost more in staff time than it saves.
Want one of these jobs running on your own data, with the cost measured? See the fixed-price AI pilot