فا

Rankings of models

Which models process the most tokens across the global market, each one’s share, and how it moved against the previous period.

Last day: Oct 30, 2026

By usage across the global model market, not IranRouter’s traffic alone.

The top three

DeepSeek: DeepSeek V4.1 Flash 1017DeepSeek

23.6%Share of tokens

No change in 7 days

Not in the IranRouter catalogue

Z.ai: GLM 5.3 FlashZ.ai

8.3%Share of tokens

Up 2 places on last week

One message on IranRouter: 292 IRR

Daily share over 90 days

The window’s eight most used; the rest under “Other”.

DeepSeek: DeepSeek V4.1 Flash 1017Z.ai: GLM 5.3 FlashDeepSeek: DeepSeek V4.1 FlashOpenAI: GPT-6 LunaTencent: Hy4 previewAnthropic: Claude Opus 5.5StepFun: Step 5 PreviewMiniMax: MiniMax M3.1 FlashOther

Biggest climbers

  1. MoonshotAI: Kimi K3.1 ProUp 35 places on last week1.1%
  2. StepFun: Step 5 PreviewUp 13 places on last week3.3%
  3. Qwen: Qwen3.7 FlashUp 12 places on last week0.18%
  4. OpenAI: GPT-6.1 SolUp 7 places on last week1.9%
  5. OpenAI: GPT-5.6 LunaUp 6 places on last week1.6%

Biggest fallers

  1. Meta: Muse Spark 1.3 ContributorDown 8 places on last week1.4%
  2. Xiaomi: MiMo-V2.6-FlashDown 7 places on last week1.6%
  3. MiniMax: MiniMax M3Down 7 places on last week0.43%
  4. OpenAI: gpt-oss-120bDown 6 places on last week0.16%
  5. MiniMax: MiniMax M3.1 FlashDown 5 places on last week2.6%

New entries

  1. OpenAI: GPT-6.1 TerraNew0.25%
  2. StepFun: Step 3.7 FlashNew0.20%
  3. Thinking Machines: InklingNew0.17%

Share by maker

Each maker’s ranked models added up; the rest of the market under “Other”.

  1. 1DeepSeekMost used: DeepSeek: DeepSeek V4.1 Flash 101735.7%
  2. 2Z.aiMost used: Z.ai: GLM 5.3 Flash11.2%
  3. 3OpenAIMost used: OpenAI: GPT-6 Luna10.1%
  4. 4AnthropicMost used: Anthropic: Claude Opus 5.56.8%
  5. 5TencentMost used: Tencent: Hy4 preview6.4%
  6. 6StepfunMost used: StepFun: Step 5 Preview3.5%
  7. 7NVIDIAMost used: NVIDIA: Nemotron 3 Ultra3.3%
  8. 8MinimaxMost used: MiniMax: MiniMax M3.1 Flash3.0%

Every ranked model

ChangeModel
1No change in 7 daysDeepSeek: DeepSeek V4.1 Flash 1017DeepSeek23.6%Not in the IranRouter catalogue
2Up 2 places on last weekZ.ai: GLM 5.3 FlashZ.ai8.3%292
3Down 1 places on last weekDeepSeek: DeepSeek V4.1 FlashDeepSeek7.9%931
4Up 4 places on last weekOpenAI: GPT-6 LunaOpenAI5.7%Not in the IranRouter catalogue
5No change in 7 daysTencent: Hy4 previewTencent4.8%7,991
6No change in 7 daysAnthropic: Claude Opus 5.5Anthropic3.8%49,284
7Up 13 places on last weekStepFun: Step 5 PreviewStepfun3.3%Not in the IranRouter catalogue
8Down 5 places on last weekMiniMax: MiniMax M3.1 FlashMinimax2.6%Not in the IranRouter catalogue
9Up 1 places on last weekZ.ai: GLM 5.3Z.ai2.4%2,299
10Up 1 places on last weekDeepSeek: DeepSeek V4 Flash 0731DeepSeek2.0%977
11Up 7 places on last weekOpenAI: GPT-6.1 SolOpenAI1.9%Not in the IranRouter catalogue
12Down 3 places on last weekNVIDIA: Nemotron 3 UltraNVIDIA1.8%6,571
13Up 1 places on last weekTypeSafe: Jev 1.13Typesafe1.6%Not in the IranRouter catalogue
14Down 7 places on last weekXiaomi: MiMo-V2.6-FlashXiaomi1.6%Not in the IranRouter catalogue
15Up 6 places on last weekOpenAI: GPT-5.6 LunaOpenAI1.6%2,738
16Down 3 places on last weekTencent: Hy3Tencent1.6%1,446
17Down 2 places on last weekDeepSeek: DeepSeek V4 Flash 0423DeepSeek1.4%747
18Down 2 places on last weekAnthropic: Claude Haiku 5.5Anthropic1.4%1,232
19Down 2 places on last weekAnthropic: Claude Sonnet 5.5Anthropic1.4%Not in the IranRouter catalogue
20Down 8 places on last weekMeta: Muse Spark 1.3 ContributorMeta1.4%821
1 DeepSeek: DeepSeek V4.1 Flash 101723.6%DeepSeekNo change in 7 days
2 Z.ai: GLM 5.3 Flash8.3%Z.aiUp 2 places on last week3 DeepSeek: DeepSeek V4.1 Flash7.9%DeepSeekDown 1 places on last week
4 OpenAI: GPT-6 Luna5.7%OpenAIUp 4 places on last week
5 Tencent: Hy4 preview4.8%TencentNo change in 7 days6 Anthropic: Claude Opus 5.53.8%AnthropicNo change in 7 days
7 StepFun: Step 5 Preview3.3%StepfunUp 13 places on last week
8 MiniMax: MiniMax M3.1 Flash2.6%MinimaxDown 5 places on last week
9 Z.ai: GLM 5.32.4%Z.aiUp 1 places on last week10 DeepSeek: DeepSeek V4 Flash 07312.0%DeepSeekUp 1 places on last week
11 OpenAI: GPT-6.1 Sol1.9%OpenAIUp 7 places on last week
12 NVIDIA: Nemotron 3 Ultra1.8%NVIDIADown 3 places on last week
13 TypeSafe: Jev 1.131.6%TypesafeUp 1 places on last week
14 Xiaomi: MiMo-V2.6-Flash1.6%XiaomiDown 7 places on last week
15 OpenAI: GPT-5.6 Luna1.6%OpenAIUp 6 places on last week16 Tencent: Hy31.6%TencentDown 3 places on last week17 DeepSeek: DeepSeek V4 Flash 04231.4%DeepSeekDown 2 places on last week18 Anthropic: Claude Haiku 5.51.4%AnthropicDown 2 places on last week
19 Anthropic: Claude Sonnet 5.51.4%AnthropicDown 2 places on last week
20 Meta: Muse Spark 1.3 Contributor1.4%MetaDown 8 places on last week

1–20 of 50

By task

What models are used for, and which lead each task — last 7 days.

Classification34% of all tokens

  1. 1TypeSafe: Jev 1.13
  2. 2TypeSafe: Jev 1.14
  3. 3OpenAI: GPT-6 Luna
  4. 4Z.ai: GLM 5.3 Flash
  5. 5DeepSeek: DeepSeek V4.1 Flash
  6. 6Perplexity: Decider V1.2
  7. 7DeepSeek: DeepSeek V4.1 Flash 1017
  8. 8DeepSeek: DeepSeek V4 Flash 0731

MethodA model’s share is its input and output tokens divided by all tokens in the window. The previous rank comes from the same-length window before.