Blog · Published · Updated
OpenAI vs Claude API: model prices and 10,000 tasks costed
Short answer
At 10,000 tasks a month with 2,000 input tokens and 500 output tokens each, the API costs from $4.50 on GPT-6 Luna and Claude Haiku 5.5 to $450 on GPT-6 Astra and Claude Fable 5.1. Sonnet 5.5 and GPT-6.1 Sol cost $90, which is €80. In the same scenario a person checking 20% of the results costs €3,490, so the number of corrections decides the choice.
How we workDetails ↓
- 01Call30 minutes with an engineer, free
- 02PrototypeA working system on your process and data, free
- 03ImplementationTarget and fixed price in the contract; 50% at signing
- 04Acceptance50% once the agreed target is met
- analysis with an engineer and a free prototype
- €0
- team in ChatGPT or Claude, up to 10 people
- from €1,150
- automation, agent or application for one process
- from €3,000
We belong to the OpenAI and Anthropic partner programmes and learn through their official courses. What this means ↓
Author: Syntalith · 4 min read
Statistics
The model API market: statistics 2026
€80
a month for 10,000 tasks on Claude Sonnet 5.5 or GPT-6.1 Sol. Checking 20% of the results at 3 minutes each costs €3,490. Scenario.
40%
According to Menlo Ventures (2025), Anthropic holds 40% of large companies' spending on language model APIs and OpenAI 27%, based on a survey of about 500 US decision-makers.
Source: Menlo Ventures, State of Generative AI in the Enterprise (2025) ↗
40%
According to Gartner (2025), over 40% of agentic AI projects will be cancelled by the end of 2027 because of costs, unclear value or weak risk controls.
14–56%
According to Anthropic (2025), AI cut task time by 14 to 56% in studies with a control group.
Source: Anthropic, Estimating AI productivity gains (2025) ↗
50%
According to the European Commission (2026), half of EU SMEs and mid-caps name high investment costs as a barrier to AI.
Source: European Commission, Flash Eurobarometer on AI (2026) ↗
Figures checked: 22 October 2026
Prices per million tokens
A token is the unit a model splits text into. Input covers the instruction, the question, the history and the material. Output is the tokens of the generated answer. OpenAI's rates come from the OpenAI API price list checked on 11 October 2026, and Anthropic's from the Claude API price list checked on 22 October 2026. Amounts are in USD, before tax, for the direct API and a short context.
2,000 input tokens and 500 output tokens per task, one attempt, no cache. ECB rate of 1.1205 USD per euro.
Źródło: Anthropic, API pricing (2026) ↗
| Model | Input, 20 million | Output, 5 million | Total USD | Total € |
|---|---|---|---|---|
| GPT-6 Luna | $2 | $2.50 | $4.50 | €4 |
| GPT-6.1 Sol | $40 | $50 | $90 | €80 |
| GPT-6 Astra | $200 | $250 | $450 | €402 |
| Claude Haiku 5.5 | $2 | $2.50 | $4.50 | €4 |
| Claude Sonnet 5.5 | $40 | $50 | $90 | €80 |
| Claude Opus 5.5 | $80 | $100 | $180 | €161 |
| Claude Fable 5.1 | $200 | $250 | $450 | €402 |
The table compares rates at identical usage. It does not rank quality or speed. The same document can come to a different number of tokens with each vendor, and answers differ in length. In a pilot, replace the assumptions with the usage counters from the vendor's console.
What changes the bill
The growing history is the easiest thing to miss. If every step of an agent sends the whole conversation and the documents, you pay for that input on every call. One ticket can take many steps.
- Without cache
- 20 million × $2 + 5 million × $10 = $90
- With cache: 5 million new input tokens × $2
- $10
- Reading 15 million tokens from cache × $0.10
- $1.50
- Output unchanged
- $50
Total $61.50 a month (€55), 32% less
Assumption: 1,500 of the 2,000 input tokens are the same instruction, and every task hits the cache. Writing to the cache with one-hour storage costs $4 per million tokens and comes on top at every refresh. Without measuring cache hits, do not apply the read rate to all traffic. Euros at the ECB rate of 1.1205 USD per euro on 22 October 2026.
Batch mode lowers the cost of tasks that do not need an answer straight away. Count paid search, file storage and the app's infrastructure separately.
The cost of checking decides the choice
Scenario: of 10,000 cases, 20% need three minutes of checking, which is 100 hours. At €34.90 an hour, the EU average hourly labour cost according to Eurostat, that is €3,490 a month. Sonnet 5.5 tokens cost €80 over the same period.
- Sonnet 5.5
- €80 API + €3,490 checking (20% of cases) = €3,570
- Opus 5.5
- €161 API + €1,745 checking (10% of cases) = €1,906
- Cost of an accepted case
- €0.36 on Sonnet, €0.19 on Opus
The dearer model saves €1,665 a month, if it really halves the corrections
The share of cases to check and the checking time are the scenario's assumptions, which you measure in a pilot on the same set of cases. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat. The API at the ECB rate of 1.1205 USD per euro on 22 October 2026.
Cost of an accepted case = (API + tools + checking + corrections) / number of accepted cases. Rejected attempts also use tokens and stay in the count. We run this count in the test on your cases before we price the build.
A cheap model with a dearer path for hard cases
You can start with a cheaper model and pass hard cases to a stronger one. If the first attempt costs $0.004 and the second $0.02, a case passed on costs $0.024 for both calls. The condition is a rule that recognises an incomplete result. The confidence the model expresses is not enough.
How to choose a model for the build
- 01SetThe same typical cases and exceptions for every model.
- 02CriteriaCorrectness set before scoring, the model version and settings recorded.
- 03MeasurementResponse time, tokens, retries and correction time.
- 04CountThe cost of an accepted case for each model.
- 05CheckThe same measurement on a new batch of data.
GPT-6.1 Sol and Sonnet 5.5 have the same rates, but usage and the share of corrections can differ. Before you change models, check compatibility of requests, tools and streaming. The guide to migrating to Sonnet 5.5 describes the changes from the model documentation. Ready-made products are compared in Claude or ChatGPT for business.
