What AI tokens really cost, per million.

Subscriptions, decoded

5-hour windows, weekly caps and credit pools, turned into real tokens and a price per 1M at full use. At full use, the coding plans below cost $0.021 to $0.10 per 1M tokens. The same agentic workload through the Claude Sonnet 5.5 API costs $1.05: 10 to 51 times more.

How it stays true

Four steps for every number. Step 1 is done for the seed index; the daily re-check and the review are being built.

1. 1

   Sourced

   Every price links to the provider's own page. If a provider doesn't publish a number, the index says “unknown”. Nothing is guessed.

   GLM-5.3 on api.z.ai: $1.40 in, $4.40 out, from docs.z.ai

2. 2

   Re-checked daily

   Each of the 158 sources has a written fetch recipe. Scheduled agents will re-read every provider once a day, and a missed check will mark the row stale.

   Planned. Not running yet.

3. 3

   Reviewed change

   When a page says something new, the agent will open a change. A person reviews it before the new price goes live.

   Planned, as pull requests on the open dataset.

4. 4

   Checked stamp

   Each number shows the date it was last checked against its source, everywhere it appears.

   $1.40 per 1M input, checked 5 Oct 2026

For developers and agents

Today: the whole dataset as JSON and CSV on GitHub, and llms.txt. Planned: a free JSON API and an MCP server. The responses below are built from today's data.

$ curl https://tokker.dev/api/v1/prices?model=claude-opus-5.5

{
  "model": "claude-opus-5.5",
  "unit": "usd_per_1m_tokens",
  "offers": [
    {
      "provider": "anthropic",
      "input": 4.0,
      "output": 20.0,
      "cache_read": 0.2,
      "batch_discount_pct": 50,
      "source_url": "https://platform.claude.com/docs/en/about-claude/pricing.md",
      "last_verified_at": "2026-10-05"
    },
    {
      "provider": "google-vertex",
      "input": 4.0,
      "output": 20.0,
      "cache_read": 0.2,
      "source_url": "https://cloud.google.com/vertex-ai/generative-ai/pricing",
      "last_verified_at": "2026-10-05"
    }
  ]
}

{
  "mcpServers": {
    "tokker": { "url": "https://tokker.dev/mcp" }
  }
}

// tool call
cheapest_provider({ "model": "glm-5.3" })

// result
{
  "provider": "deepinfra",
  "input": 0.9,
  "output": 4.0,
  "source_url": "https://api.deepinfra.com/models/list",
  "last_verified_at": "2026-10-05"
}