Sources and methodology
How we find things out, and how you can check us.
What makes a topic
We look for questions practitioners are actually asking, in one of the three subjects we cover, where the existing answers leave something out. The reason each topic was chosen is recorded.
How claims are sourced
We read primary sources first: vendor pricing pages, documentation, changelogs, official announcements and research. Every claim in an entry is linked to the page it comes from and checked against that page before publication. Claims the source does not support are rewritten or cut, and a draft that cannot be supported is not published.
Estimates
Sometimes a reader needs a number no one publishes, usually a price. Where an entry gives an informed estimate, it is labelled as an estimate in the sentence itself and says what it is based on.
How to check us
Every claim in an entry is linked to its source, and the sources are listed with each entry. If a link no longer says what we said it says, tell us at ops@3ammarketer.com.
Measurements we make ourselves
Where an entry reports a measurement of our own, the entry states how it was made: the setup, sample size and date. When a test was narrow, the entry says so next to the finding rather than in a footnote. We do not claim hands-on use of products.
Our data sets
When we publish a data set, it comes with its method and changelog, and it is downloadable with an attribution licence, so you can check our working or use the data yourself.
We are building our own data sets on AI tools, models and marketing costs. Each one is listed here once it is live, with its method and a download.
LLM API pricing timeline: input, output and cached input prices per million tokens for the main LLM APIs, read daily from each provider's official pricing page, with every change dated. Download (CSV, CC BY 4.0).
Open model release tracker: open-weight model releases from the main publishers, read daily from the public Hugging Face models API, with release date, parameter count and licence from each publisher's own repository. Download (CSV, CC BY 4.0).
AI tool pricing tracker: monthly list prices of AI writing, SEO, coding and agent tools by plan, with annual billing, seat basis and the main usage limit, read daily from each vendor's public pricing page, with every change dated. Download (CSV, CC BY 4.0).
AI launch signals: how many AI launch items our news monitoring read in the last 7 and 30 days, by source type, official announcement site and day, and which announcements the most sources carried. Counts only, with headlines linking to the original pages. Download daily counts (CSV, CC BY 4.0).
How we collect our data
Prices come from each seller's own official pricing page, never from third-party price lists or press coverage. Open model details come from each publisher's own Hugging Face repository, through the public Hugging Face models API.
Every source is read once a day. A price is recorded only when it differs from the last one we hold: each change becomes a new dated row and the old price stays in the history, so you can see what a plan or model cost on any past date. When a plan disappears from a vendor's page, it is marked as removed from that page, not deleted.
The Publisher reads every entry and corrects or removes it if needed. If you spot a wrong value, write to ops@3ammarketer.com with the page and the figure; corrections are logged at /corrections.
All data sets are free to reuse under CC BY 4.0: credit us and link to the data set page.
Model API prices: sources
- Anthropic: docs.anthropic.com/en/docs/about-claude/pricing, platform.claude.com/docs/en/about-claude/pricing, claude.com/pricing and 1 more pages. Last change recorded .
- DeepSeek: api-docs.deepseek.com/quick_start/pricing. Last change recorded .
- Google Gemini: ai.google.dev/gemini-api/docs/pricing, ai.google.dev/gemini-api/docs/pricing?hl=en. Last change recorded .
- Meta: llama.developer.meta.com/docs/pricing-rate-limits. Last change recorded .
- Mistral: docs.mistral.ai/models/mistral-large-3-25-12, docs.mistral.ai/models/mistral-medium-3-5-26-04, docs.mistral.ai/models/mistral-small-4-0-26-03 and 4 more pages. Last change recorded .
- Moonshot: platform.kimi.ai/docs/pricing/chat.md, platform.kimi.ai/docs/pricing/chat. Last change recorded .
- OpenAI: platform.openai.com/docs/pricing. Last change recorded .
- OpenRouter: openrouter.ai/unbiased/pareto-26.9, openrouter.ai/liquid/lfm-2.5-24b-a2b, openrouter.ai/google/gemini-4-argon and 363 more pages. Last change recorded .
- Qwen: alibabacloud.com/help/en/model-studio/model-pricing. Last change recorded .
- xAI: docs.x.ai/developers/models.md. Last change recorded .
AI tool plan prices: sources
- ChatGPT: help.openai.com/en/articles/6950777-what-is-chatgpt-plus, help.openai.com/en/articles/9793128-what-is-chatgpt-pro, help.openai.com/en/articles/8792536-what-is-chatgpt-business. Last change recorded .
- Claude: claude.com/pricing. Last change recorded .
- Clearscope: clearscope.io/pricing. Last change recorded .
- Copy.ai: copy.ai/prices. Last change recorded .
- Cursor: cursor.com/pricing. Last change recorded .
- ElevenLabs: elevenlabs.io/pricing. Last change recorded .
- Frase: frase.io/pricing. Last change recorded .
- GitHub Copilot: github.com/features/copilot/plans. Last change recorded .
- Google AI: one.google.com/intl/en_us/about/google-ai-plans. Last change recorded .
- Grok (X Subscription): docs.x.ai/x-subscription/pricing. Last change recorded .
- Gumloop: gumloop.com/pricing. Last change recorded .
- HeyGen: heygen.com/pricing. Last change recorded .
- Jasper: jasper.ai/pricing. Last change recorded .
- Lindy: lindy.ai/pricing. Last change recorded .
- Lovable: lovable.dev/pricing. Last change recorded .
- Microsoft 365 Copilot: microsoft.com/en-us/microsoft-365/copilot/pricing. Last change recorded .
- Microsoft 365 Premium: microsoft.com/en-us/store/b/copilotpro. Last change recorded .
- Midjourney: docs.midjourney.com/hc/en-us/articles/27870484040333-Comparing-Midjourney-Plans. Last change recorded .
- Notion: notion.com/pricing. Last change recorded .
- Otter.ai: otter.ai/pricing. Last change recorded .
- Perplexity: perplexity.ai/pro. Last change recorded .
- Replit: replit.com/pricing. Last change recorded .
- Surfer: surferseo.com/pricing. Last change recorded .
- Windsurf: windsurf.com/pricing. Last change recorded .
- Writesonic: writesonic.com/pricing. Last change recorded .
- Zapier: zapier.com/pricing. Last change recorded .
Open models: sources
The Hugging Face repositories of meta-llama, mistralai, Qwen, deepseek-ai, google, microsoft, nvidia, allenai, ibm-granite, CohereLabs, moonshotai, zai-org, openai, HuggingFaceTB, tiiuae, MiniMaxAI, LiquidAI, ai21labs, read daily.
Our worked examples
Where an entry gives "our worked example", the figures are plain arithmetic done in code on numbers the cited sources state, such as a published fee rate applied to round order values, or the change between an old and a new published price. The inputs are given next to the result, and any order values are illustrative, not data.
Case studies
Some articles are case studies from engagements the Publisher worked on. He supplies the case document; the company, client, products and people are not named, and any detail that would identify them is replaced with a neutral description. Every figure in the article is checked against that source document before it can publish, and the Publisher approves each case study himself. Because the company is not named, you cannot verify these figures independently, and each case study says so near the top.
Corrections
When a published fact turns out to be wrong or out of date, it is fixed and the change is logged at /corrections.