Skip to content

Live usage

LLM Rankings

Token usage across zrouter, by model and by author.

Tokens in range

54B+3%vs previous period

Data to Oct 16, 2026, 08:21 WIB · since Sep 3, 2026 · refreshes every 60s

Top models

Tokens per day. Top 8 models, the rest in Other.

  • DeepSeek V4 Flash
  • Gemini 3.8 Flash
  • DeepSeek V4 Pro
  • Claude Haiku 5.5
  • GLM 5.3 Flash
  • GLM 5.3
  • DeepSeek V4.1 Flash
  • GPT-6 Luna
  • Other
  • Partial period

Leaderboard

  1. 1.
    23.5B tokens0%
  2. 2.
    Claude Haiku 5.5by Anthropic
    7.8B tokensNew
  3. 3.
    6.8B tokens−21%
  4. 4.
    GPT-6 Lunaby OpenAI
    6.5B tokens−10%
  5. 5.
    GLM 5.3by Z.ai
    2.3B tokens−48%
  6. 6.
    1.1B tokens−68%
  7. 7.
    1.1B tokens−17%
  8. 8.
    DeepSeek V4 Proby DeepSeek
    1B tokens+296%
  9. 9.
    MiniMax M3by MiniMax
    992.4M tokens+22%
  10. 10.
    GPT 6.1 Solby OpenAI
    775.9M tokens+188%
  1. 11.
    502.7M tokens−47%
  2. 12.
    Claude Sonnet 5.5by Anthropic
    382.1M tokensNew
  3. 13.
    GPT-6 Solby OpenAI
    256M tokens+34%
  4. 14.
    Kimi K3by Moonshot AI
    208M tokens+302%
  5. 15.
    Claude Opus 4.6by Anthropic
    196.6M tokens−37%
  6. 16.
    175.3M tokens−36%
  7. 17.
    Claude Opus 5by Anthropic
    58.7M tokens−72%
  8. 18.
    GPT-6 Astraby OpenAI
    41.7M tokens+18%
  9. 19.
    GPT 5.6 Lunaby OpenAI
    24.3M tokens−58%
  10. 20.
    GLM 5.2by Z.ai
    10.3M tokens+129%

Market share by author

  • DeepSeek47.56%
  • Z.ai16.97%
  • Anthropic15.75%
  • OpenAI14.16%
  • Google2.98%
  • MiniMax1.84%
  • Moonshot AI0.39%
  • Other0.35%

Last 7 days vs the 7 before.

  1. 1.
    Claude Haiku 5.5by Anthropic
    7.8BNew
  2. 2.
    Claude Sonnet 5.5by Anthropic
    382.1MNew
  3. 3.
    Claude Opus 4.5by Anthropic
    4.8MNew
  4. 4.
    Claude Opus 5 5by Anthropic
    3.4MNew
  5. 5.
    GPT 5.6 Terraby OpenAI
    1M>999%
  6. 6.
    Kimi K3by Moonshot AI
    208M+302%
  7. 7.
    DeepSeek V4 Proby DeepSeek
    1B+296%
  8. 8.
    GPT 5.6 Solby OpenAI
    5.3M+191%
  9. 9.
    GPT 6.1 Solby OpenAI
    775.9M+188%
  10. 10.
    GLM 5.2by Z.ai
    10.3M+129%

Fastest

Median output speed of streamed requests, last 7 days.

  1. 1.
    138 tok/sTTFT 2.3 s
  2. 2.
    103 tok/sTTFT 16.4 s
  3. 3.
    89 tok/sTTFT 2.5 s
  4. 4.
    84 tok/sTTFT 25.6 s
  5. 5.
    Claude Sonnet 5.5by Anthropic
    64 tok/sTTFT 5.9 s
  6. 6.
    Claude Haiku 5.5by Anthropic
    51 tok/sTTFT 4.6 s
  7. 7.
    Claude Opus 5by Anthropic
    46.5 tok/sTTFT 5.6 s
  8. 8.
    DeepSeek V4 Proby DeepSeek
    43 tok/sTTFT 2.3 s
  9. 9.
    43 tok/sTTFT 2 s
  10. 10.
    Claude Opus 4.6by Anthropic
    39 tok/sTTFT 5.2 s
Data tables
Leaderboard
RankModelAuthorTokensChange
1DeepSeek V4.1 FlashDeepSeek23,517,700,0390%
2Claude Haiku 5.5Anthropic7,846,981,717New
3GLM 5.3 FlashZ.ai6,809,555,398−21%
4GPT-6 LunaOpenAI6,538,667,109−10%
5GLM 5.3Z.ai2,339,020,244−48%
6DeepSeek V4 FlashDeepSeek1,127,848,398−68%
7Gemini 3.8 FlashGoogle1,104,999,979−17%
8DeepSeek V4 ProDeepSeek1,019,520,114+296%
9MiniMax M3MiniMax992,351,832+22%
10GPT 6.1 SolOpenAI775,856,595+188%
11Gemini 3.7 FlashGoogle502,723,570−47%
12Claude Sonnet 5.5Anthropic382,148,297New
13GPT-6 SolOpenAI256,041,101+34%
14Kimi K3Moonshot AI207,972,032+302%
15Claude Opus 4.6Anthropic196,618,296−37%
16Qwen 3.8 FlashQwen175,339,016−36%
17Claude Opus 5Anthropic58,744,269−72%
18GPT-6 AstraOpenAI41,725,995+18%
19GPT 5.6 LunaOpenAI24,252,162−58%
20GLM 5.2Z.ai10,307,068+129%
21Grok 4.7xAI8,019,521+22%
22Claude Fable 5Anthropic5,885,103+65%
23GPT 5.6 SolOpenAI5,263,994+191%
24Claude Opus 4.5Anthropic4,841,855New
25Claude Opus 5 5Anthropic3,368,605New
26Qwen 3.8 MaxQwen2,447,986−94%
27GPT 5.6 TerraOpenAI1,046,357>999%
28HY4 PreviewTencent527,618−98%
29Gemini 3.1 ProGoogle45,127+666%
30Kimi K2.7 CodeMoonshot AI474−100%
Tokens by model per period
PeriodDeepSeek V4 FlashGemini 3.8 FlashDeepSeek V4 ProClaude Haiku 5.5GLM 5.3 FlashGLM 5.3DeepSeek V4.1 FlashGPT-6 LunaOtherTotal
Oct 9, 2026 (Partial period)283,812,132147,314,075104,567,0040973,846,032291,419,2002,312,069,319857,629,973543,772,8135,514,430,548
Oct 10, 2026174,849,96069,725,940114,713,1310508,598,976369,371,6183,400,682,7791,135,784,071609,098,4116,382,824,886
Oct 11, 202636,570,863102,782,184163,961,7142,517,056,7611,331,093,784801,501,6853,804,689,5801,221,466,593581,795,61910,560,918,783
Oct 12, 202631,978,440115,457,305263,454,9552,214,594,6141,006,994,044512,423,9343,151,229,967919,716,306501,286,3668,717,135,931
Oct 13, 2026206,375,038150,715,608239,054,079509,617,6261,292,027,760155,213,7232,743,269,634893,611,729350,198,8186,540,084,015
Oct 14, 2026301,113,597278,179,10952,458,0191,278,948,892857,502,92987,871,5384,162,158,931891,112,764402,560,6268,311,906,405
Oct 15, 202693,083,046197,991,58869,262,0701,316,880,507723,135,552108,189,7993,364,140,965476,247,360614,658,3656,963,589,252
Oct 16, 2026 (Partial period)65,32242,834,17012,049,1429,883,317116,356,32113,028,747579,458,864143,098,31352,155,855968,930,051
Market share by author
AuthorTokensShare
DeepSeek25,665,068,55147.56%
Z.ai9,158,882,71016.97%
Anthropic8,498,588,14215.75%
OpenAI7,642,853,31314.16%
Google1,607,768,6762.98%
MiniMax992,351,8321.84%
Moonshot AI207,972,5060.39%
Qwen177,787,0020.33%
xAI8,019,5210.02%
Tencent527,6180.00%
Trending models
ModelAuthorTokensChange
Claude Haiku 5.5Anthropic7,846,981,717New
Claude Sonnet 5.5Anthropic382,148,297New
Claude Opus 4.5Anthropic4,841,855New
Claude Opus 5 5Anthropic3,368,605New
GPT 5.6 TerraOpenAI1,046,357>999%
Kimi K3Moonshot AI207,972,032+302%
DeepSeek V4 ProDeepSeek1,019,520,114+296%
GPT 5.6 SolOpenAI5,263,994+191%
GPT 6.1 SolOpenAI775,856,595+188%
GLM 5.2Z.ai10,307,068+129%
Fastest models
ModelAuthorSpeedTTFTSamples
DeepSeek V4 FlashDeepSeek138 tok/s2.3 s1k+
Gemini 3.7 FlashGoogle103 tok/s16.4 s1k+
DeepSeek V4.1 FlashDeepSeek89 tok/s2.5 s10k+
Gemini 3.8 FlashGoogle84 tok/s25.6 s10k+
Claude Sonnet 5.5Anthropic64 tok/s5.9 s1k+
Claude Haiku 5.5Anthropic51 tok/s4.6 s10k+
Claude Opus 5Anthropic46.5 tok/s5.6 s200+
DeepSeek V4 ProDeepSeek43 tok/s2.3 s10k+
Qwen 3.8 FlashQwen43 tok/s2 s1k+
Claude Opus 4.6Anthropic39 tok/s5.2 s1k+

How it is counted

Input plus output tokens of every request (cache reads excluded), grouped by Asia/Jakarta day. Models opted out of public rankings count in Other. Image models are not included.

Rankings API