Skip to content
FreeAIDevTools

How we verify

Updated

Every price and free-tier limit on this site is read from the provider's own pricing page or docs, dated, and re-checked every month. When a provider does not publish a number, we say so instead of filling the gap with a guess.

Model prices

  1. Open each provider's official pricing page (OpenAI, Anthropic, Google AI Studio and Vertex, xAI, DeepSeek, Mistral) and read input, output and cached-input prices per million tokens, plus the context window from the model docs.
  2. Compare against OpenRouter's public model list, which flags drift early. When the two disagree, the provider's own page wins, because OpenRouter sometimes shows marketplace or off-peak rates.
  3. Note promotions with their end dates, and move models the provider has shut down out of the tools.
  4. Save a dated snapshot of the whole dataset and log what changed in the changelog.

The pricing dataset covers 50+ models from 7 providers. Last full check: November 2026. See exactly what changed on the changelog.

Free-tier limits

  1. Read the provider's own rate-limit and pricing docs for requests per minute, per day and per month, token caps, and whether a card is required.
  2. Read the terms for the two things that matter most on a free plan: whether your prompts can be used for training, and whether commercial use is allowed.
  3. Record the source link and the check date on the row. Anything not published by the provider shows as "Not confirmed". Some providers, such as Google and Mistral, only show free limits inside your logged-in console, and we say that rather than copy numbers from third-party blogs.
  4. Re-check every row each month and log changes with a source.

15 providers are tracked today, each checked on or after 29 Oct 2026. See them on Free LLM APIs. As we add more categories, they appear under free tiers by category.

Token counts

OpenAI counts are exact: the token counter runs OpenAI's own tiktoken tokenizer in your browser. We only run OpenAI's tokenizer locally, so counts for other providers are estimated from characters-per-token ratios. For exact counts, use the provider's own counting endpoint (Anthropic's count_tokens or Gemini's countTokens).

Corrections

If you spot a number that is out of date, use the contact page with a link to the source. Confirmed fixes ship the same week and are noted in the changelog.