What a job actually costs on each plan
thecontinuitylayer.com/map/real-costsChecked 21 October 2026. US dollars. Prices from each provider's own pricing page.
There are two ways to pay for AI. A plan is a flat monthly price with usage limits. The API charges for every token you send and get back. This page shows both for a few common jobs, using only the prices the providers publish. For the API, we show the tokens we assumed and the math. We make no recommendations and use no affiliate links.
Plans: one monthly price
On a plan, the writing, PDF and coding jobs below cost nothing extra, as long as you stay within the plan's usage limits. Limits differ by plan and are not priced per job.
- Free: $0. (Source: ChatGPT pricing, checked 21 October 2026)
- Go: $8 a month. US price; priced locally in some countries. (Source: Introducing ChatGPT Go, checked 21 October 2026)
- Plus: $20 a month. (Source: ChatGPT pricing, checked 21 October 2026)
- Pro: $100, $200 or $500 a month. Three Pro tiers with different usage limits. (Source: ChatGPT release notes, checked 21 October 2026)
- Free: $0. (Source: Claude pricing, checked 21 October 2026)
- Pro: $20 a month. $17 a month if paid yearly ($200 up front). (Source: Claude pricing, checked 21 October 2026)
- Max: From $100 a month. (Source: Claude pricing, checked 21 October 2026)
- Free: $0. (Source: Google AI plans, checked 21 October 2026)
- Google AI Plus: $4.99 a month. (Source: Google AI plans, checked 21 October 2026)
- Google AI Pro: $19.99 a month. (Source: Google AI plans, checked 21 October 2026)
- Google AI Ultra: From $99.99 a month. (Source: Google AI plans, checked 21 October 2026)
API prices: per million tokens
Standard rates, before any batch or caching discount. A token is a piece of a word.
| Model | In | Out | Source |
|---|---|---|---|
| Claude Opus 5.5 (Anthropic) | $4 | $20 | Claude API pricingchecked 21 October 2026 |
| Claude Sonnet 5.5 (Anthropic) | $2 | $10 | Claude API pricingchecked 21 October 2026 |
| Claude Haiku 5.5 (Anthropic)Prompts over 100,000 tokens: $0.50 in, $2.50 out. | $0.10 | $0.50 | Claude API pricingchecked 21 October 2026 |
| GPT-6 Astra (OpenAI)OpenAI's current flagship API model. | $10 | $50 | OpenAI modelschecked 21 October 2026 |
| GPT-5.4 (OpenAI)Prompts over 272,000 tokens: $5 in, $22.50 out. | $2.50 | $15 | GPT-5.4 model pagechecked 21 October 2026 |
| Gemini 3.1 Pro (preview) (Google)Prompts over 200,000 tokens: $4 in, $18 out. | $2 | $12 | Gemini API pricingchecked 21 October 2026 |
| Gemini 3.6 Flash (Google)Google lists these rates through 31 December 2026, then $1.50 in and $7.50 out. | $0.75 | $3.75 | Gemini API pricingchecked 21 October 2026 |
Write a 1,000-word draft
What we assumed: One request. Your instructions: about 300 tokens in. The draft: about 1,350 tokens out (we assume about 0.75 words per token).
- Claude Opus 5.5300 in × $4/M + 1,350 out × $20/M$0.028
- Claude Sonnet 5.5300 in × $2/M + 1,350 out × $10/M$0.014
- Claude Haiku 5.5300 in × $0.10/M + 1,350 out × $0.50/M$0.0007
- GPT-6 Astra300 in × $10/M + 1,350 out × $50/M$0.070
- GPT-5.4300 in × $2.50/M + 1,350 out × $15/M$0.021
- Gemini 3.1 Pro (preview)300 in × $2/M + 1,350 out × $12/M$0.017
- Gemini 3.6 Flash300 in × $0.75/M + 1,350 out × $3.75/M$0.0053
Summarize a 50-page PDF
What we assumed: One request. 50 pages at about 2,000 tokens a page, plus a 200-token instruction: 100,200 tokens in. A one-page summary: about 800 tokens out. Anthropic says a PDF page typically uses 1,500 to 3,000 tokens, depending on how dense it is. (Source: Claude PDF support, checked 21 October 2026)
- Claude Opus 5.5100,200 in × $4/M + 800 out × $20/M$0.417
- Claude Sonnet 5.5100,200 in × $2/M + 800 out × $10/M$0.208
- Claude Haiku 5.5100,200 in × $0.50/M + 800 out × $2.50/M (long-prompt rates)$0.052
- GPT-6 Astra100,200 in × $10/M + 800 out × $50/M$1.04
- GPT-5.4100,200 in × $2.50/M + 800 out × $15/M$0.263
- Gemini 3.1 Pro (preview)100,200 in × $2/M + 800 out × $12/M$0.210
- Gemini 3.6 Flash100,200 in × $0.75/M + 800 out × $3.75/M$0.078
A day of coding help
What we assumed: 40 requests in a day, each sending about 6,000 tokens of code and question and getting about 800 tokens back: 240,000 in and 32,000 out. No prompt caching or batch discounts. Coding agents that read whole projects can use many times more.
- Claude Opus 5.5240,000 in × $4/M + 32,000 out × $20/M$1.60
- Claude Sonnet 5.5240,000 in × $2/M + 32,000 out × $10/M$0.800
- Claude Haiku 5.5240,000 in × $0.10/M + 32,000 out × $0.50/M$0.040
- GPT-6 Astra240,000 in × $10/M + 32,000 out × $50/M$4.00
- GPT-5.4240,000 in × $2.50/M + 32,000 out × $15/M$1.08
- Gemini 3.1 Pro (preview)240,000 in × $2/M + 32,000 out × $12/M$0.864
- Gemini 3.6 Flash240,000 in × $0.75/M + 32,000 out × $3.75/M$0.300
Generate 20 images
Image prices below are API prices, per image. Plans do not price images one by one.
- Gemini 3.1 Flash Image (Google)20 × $0.067Google lists about $0.067 for each 1K-resolution image (about $0.101 at 2K and $0.151 at 4K). (Source: Gemini API pricing, checked 21 October 2026)$1.34
- gpt-image models (OpenAI)OpenAI prices images by tokens, and the cost depends on size and quality. We could not confirm a current per-image figure on its pricing page, so we show none. (Source: OpenAI API pricing, checked 21 October 2026)Not shown
- Claude (Anthropic)Anthropic's pricing page lists no image-generation price. (Source: Claude pricing, checked 21 October 2026)Not shown
Run a model on your own laptop
Ollama says running models on your own computer is free. You pay for the computer and its electricity, which we do not estimate. Ollama's paid plans are for its cloud models, not local ones. (Source: Ollama pricing, checked 21 October 2026)
Prices change often; each figure carries the day we checked it. For which tool fits which job, see Which AI for which job. For dated price changes, see Did it change?. If something here is out of date, tell us through Work with Adam.