Skip to content
OfferSign up and get $10 of credit for any model in Chat and the API. Top up from $1, with no processing fees. See free models →·Some free models are at the daily cap — refills 05:30 IST; requests auto-route to a live backup. Status →
One AI API Key
Get API key

AI model rankings: the most used LLMs

Live · updated every 5 minData through Nov 14, 05:10 UTC

Real tokens served through One AI API Key. Ranked by usage, not quality.

Free right now (100 requests a day per account, no card): qwen3.8-27b:free, qwen3.8-27b-1:free, stepaudio-3-chat-preview:free

Tokens per day, by model

Last 7 days, UTC days

Hover, tap or drag across any day, or focus the chart and use the arrow keys, to see how every model did that day.

Last 7 days, rolling · to 05:10 UTC

Tokens
1.1B
26.5K requests
Models ranked
28
top 3 hold 58.2%
Open-weight share
66.4%
closed: Claude, GPT, Gemini, Grok, Muse
People
503
avg 2.3M tokens each

Most used modelsLast 7 days

Movers · last 24 hours

Rank by
Sort

Rows are focusable: press Enter for the model's details, or C to add it to the comparison of up to three models.

Tokens · share of the window

  1. 1
    qwen3.8-max-1:free
    141 peopleHot
    271M tokens
    23.6%
  2. 2
    qwen3.8-max:free
    188 peopleHot
    208M tokens
    18.1%
  3. 3
    claude-opus-5-5:free
    by anthropic222 peopleHot
    189M tokens
    16.4%
  4. 4
    claude-sonnet-5-5:free
    by anthropic151 people
    105M tokens
    9.2%
  5. 5
    glm-5.3-flash:free
    by z.ai99 peopleSpike
    56.4M tokens
    4.9%
  6. 6
    gpt-6-sol:free
    55 people
    44.2M tokens
    3.8%
  7. 7
    deepseek-v4-pro:free
    42 people
    36.3M tokens
    3.2%
  8. 8
    deepseek-v4-flash-0731:free
    30 people
    34.2M tokens
    3%
  9. 9
    qwen3.8-27b-abliterated-cyber:free
    31 people
    27M tokens
    2.3%
  10. 10
    claude-fable-5-1:free
    by anthropic104 people
    24.7M tokens
    2.2%
  11. 11
    qwen3.8-27b:free
    57 people
    19.5M tokens
    1.7%
  12. 12
    deepseek-v4-flash:free
    40 people
    16.5M tokens
    1.4%
  13. 13
    kimi-k3:free
    by moonshot ai39 people
    11.3M tokens
    1%
  14. 14
    glm-5.3:free
    by z.ai42 peopleSpike
    9.9M tokens
    0.9%
  15. 15
    qwen3.8-27b-abliterated-cyber
    11 people
    8.1M tokens
    0.7%
  16. 16
    uncens-code-glm-5.3-flash
    by unfiltered18 people
    5.2M tokens
    0.4%
  17. 17
    uncens-code-pro
    by unfiltered16 people
    3.5M tokens
    0.3%
  18. 18
    glm-5.2:free
    13 people
    3.5M tokens
    0.3%
  19. 19
    uncens-code-glm-5.3-flash:free
    by unfiltered20 people
    3.1M tokens
    0.3%
  20. 20
    claude-fable-5.2:free
    12 people
    2.5M tokens
    0.2%

How it works

Raw dataCC BY 4.0Compare prices

Quality or popularity?

Popularity. Models are ranked by the tokens they served on this gateway, not by how good they are.

How fresh is the data?

Updated every 5 minutes at the latest, and usually within seconds of a request finishing, straight from the gateway logs. The stamp at the top shows the instant the numbers are exact for. In a quiet spell with no new request the figures do not change, so the stamp stays and the line under the title says when the numbers were last checked again.

Can a model pay to rank higher?

No. Rank follows the numbers only. Nothing is sold, boosted or placed by hand.

Which version is a row?

The exact API id, in lowercase, as customers send it. x and x:free are separate rows, and so are claude-opus-5-5 and claude-opus-5.5.

Counted

Exact counts, never estimates: successful requests as the gateway logged them, by exact id. Tokens are input + output (plus cache reads and writes for Anthropic-style calls). A request counts when it finishes. Every account counts, deleted ones too. An account using two ids counts once under each. Search, fetch and code-run calls are in the totals, not in rows.

Ranking bar

A model gets its own row from 100K tokens in the window, judged for each window on its own. Smaller models stay in the totals and in the "other" row. Totals only: no ids, keys or prompts.

Windows

The totals and the list use rolling windows (24 hours, 7 days, 30 days), counted to the second up to the last update. The chart’s bars are UTC calendar days, and today’s bar is 00:00 UTC to the last update, so “last 24 hours” and “today” are different numbers. From 30 Oct 2026.

Speed

Output tokens per second, request-weighted, over requests of 1s or more. It includes time to first token, so reasoning models read low. Needs 5 timed requests.

Spike and Hot

Spike: new, or added 250K tokens (1M in longer windows) and grew 50%+ against the period before. Hot: top 3 by tokens in the window.