LastAI cover
Sign in

Usage rankings

Compare models and apps by OpenRouter token volume over the last 7 days. Tokens measure workload, not the number of users or model quality.

Nov 5–11 · OpenRouter data

Usage over seven days

Ranked by what actually flowed through OpenRouter, where apps pick among many AIs.

AI tokens in seven days207Ttokens-0.4% vs last weekTokens are the pieces of text an AI reads and writes.

Most-used models

More
  1. 1DeepSeek V4.1 Flash 1017DeepSeek25.7%down 8% from last week
  2. 2Hy4Tencent9.8%
  3. 3DeepSeek V4.1 FlashDeepSeek7.9%up 1% from last week
  4. 4GLM 5.3 FlashZ.ai5.8%down 25% from last week
  5. 5GPT-6 LunaOpenAI3.5%down 20% from last week
  6. 6Step 5 PreviewStepFun3.3%up 2% from last week
  7. 7Claude Opus 5.5Anthropic3.2%down 2% from last week
  8. 8Crouton Alphastealth2.2%
  9. 9Kimi K3.1 ProMoonshot AI2.1%up 32% from last week
  10. 10GPT-6.1 SolOpenAI2.0%up 1% from last week

Percentages show shares of all tokens. Bars are scaled to the leading model.

Token share by maker

More
  • DeepSeek36.9%
  • Tencent10.9%
  • Z.ai8.1%
  • OpenAI7.3%
  • Anthropic6.1%
  • Everyone else30.8%

Usage through OpenRouter only, not overall market share. Model values add only days in the daily top 50, so they can run slightly low.

Compare all models
Fusion · Beta

Compare multiple models and fuse one answer

Multi-model fusion is a routing method, not a ranking dataset. It combines a model panel with a judge, and cost varies by panel, so it is not treated as a LastAI score or fixed price.

Read the explainer

Scope and sources

LastAI uses documented stable APIs only. The 10K context bucket, tool-using requests, and image-input activity are shown as observed token floors. Korean or Python-specific tags, tool-call counts, and image counts lack stable APIs, so proxies are separated and never used in a composite score.

Source: OpenRouter (openrouter.ai/rankings), as of 2026-11-12T05:20:08.915Z.

Methodology · external sources

You

My accountPast forecastsMy wallet