Model rankings

Benchmarks tell you what a model can do; traffic tells you what teams actually ship. This board ranks the catalog by token volume and shows how fast each model's share is moving.

Reference data — not yet measured from live traffic

Illustrative sample data, not measurements. Every token count, share, and growth figure below is seeded from the catalog to show how this board will look. Rankings become real once the router has served live completions — the gateway needs real traffic before it can report its own.

1Sample

+22.6%

Claude Sonnet 4.5

Anthropic

884B

tokens processed · illustrative

2Sample

+148.3%

Gemini 3 Flash

Google

812B

tokens processed · illustrative

3Sample

+74.2%

GPT-5.2

OpenAI

641B

tokens processed · illustrative

Illustrative sample data — seeded from the catalog, not measured from live traffic

#	Model	Series	Tokens	Trend
1	Claude Sonnet 4.5Anthropic	Claude	884B	▲ 22.6%
2	Gemini 3 FlashGoogle	Gemini	812B	▲ 148.3%
3	GPT-5.2OpenAI	GPT	641B	▲ 74.2%
4	DeepSeek V3.2DeepSeek	DeepSeek	573B	▲ 57.4%
5	Gemini 3 ProGoogle	Gemini	486B	▲ 64.8%
6	Grok 4 FastxAI	Grok	468B	▲ 163.7%
7	Claude Opus 4.5Anthropic	Claude	402B	▲ 88.3%
8	Claude Haiku 4.5Anthropic	Claude	351B	▲ 96.4%
9	GPT-5.2 MiniOpenAI	GPT	312B	▲ 121.5%
10	Gemini 2.5 FlashGoogle	Gemini	297B	▼ 8.2%
11	GPT-5.1OpenAI	GPT	268B	▼ 18.4%
12	GPT-4o MiniOpenAI	GPT	214B	▼ 9.3%
13	Qwen3 Coder 480BAlibaba Qwen	Qwen	206B	▲ 79.6%
14	GPT-5.1 CodexOpenAI	GPT	187B	▲ 33.1%
15	gpt-oss-120bOpenAI	GPT	178B	▲ 41.6%
16	Gemini 2.5 ProGoogle	Gemini	176B	▼ 21.4%
17	Llama 3.3 70B InstructMeta	Llama	164B	▼ 6.7%
18	Kimi K2 ThinkingMoonshot AI	Kimi	158B	▲ 134.2%
19	DeepSeek R1 (0528)DeepSeek	DeepSeek	148B	▼ 16.3%
20	GPT-4oOpenAI	GPT	143B	▼ 26.8%
21	GLM 4.6Z.ai	GLM	142B	▲ 71.9%
22	Grok 4xAI	Grok	137B	▲ 28.3%
23	Gemini 2.5 Flash LiteGoogle	Gemini	129B	▲ 18.6%
24	Llama 4 MaverickMeta	Llama	121B	▲ 9.1%
25	Claude Sonnet 4Anthropic	Claude	118B	▼ 22.5%
26	Qwen3 235B A22BAlibaba Qwen	Qwen	118B	▲ 21.8%
27	Kimi K2Moonshot AI	Kimi	112B	▲ 52.1%
28	GPT-5 NanoOpenAI	GPT	96B	▲ 12.7%
29	Qwen3 MaxAlibaba Qwen	Qwen	92B	▲ 68.5%
30	o4-miniOpenAI	GPT	88B	▼ 14.1%
31	Llama 4 ScoutMeta	Llama	87B	▲ 15.3%
32	MiniMax M2MiniMax	Other	79B	▲ 112.4%
33	Llama 3.1 8B InstructMeta	Llama	73B	▼ 3.4%
34	DeepSeek V3 (0324)DeepSeek	DeepSeek	66B	▼ 24.1%
35	gpt-oss-20bOpenAI	GPT	64B	▲ 27.9%
36	GLM 4.5 AirZ.ai	GLM	61B	▲ 19.3%
37	Claude 3.7 SonnetAnthropic	Claude	57B	▼ 28.7%
38	Qwen3 32BAlibaba Qwen	Qwen	54B	▲ 11.4%
39	Mistral Medium 3Mistral AI	Mistral	52B	▲ 24.9%
40	Codestral 2508Mistral AI	Mistral	46B	▲ 17.2%
41	Claude 3.5 HaikuAnthropic	Claude	44B	▼ 11.9%
42	Claude Opus 4.1Anthropic	Claude	41B	▼ 31.2%
43	Mistral Small 3.2Mistral AI	Mistral	41B	▲ 31.7%
44	Gemma 3 27BGoogle	Other	38B	▲ 6.4%
45	Mistral Large 2Mistral AI	Mistral	34B	▼ 12.1%
46	Grok 3xAI	Grok	33B	▼ 29.5%
47	QwQ 32BAlibaba Qwen	Qwen	31B	▼ 13.6%
48	Llama 3.1 405B InstructMeta	Llama	29B	▼ 19.8%
49	Grok 3 MinixAI	Grok	27B	▼ 7.2%
50	Nova LiteAmazon	Other	26B	▲ 8.9%
51	Devstral MediumMistral AI	Mistral	24B	▲ 43.8%
52	Sonar ProPerplexity	Other	23B	▲ 16.1%
53	Command ACohere	Other	22B	▲ 4.8%
54	Nova ProAmazon	Other	19B	▼ 5.6%
55	Llama 3.1 Nemotron 70BNVIDIA	Llama	17B	▼ 10.4%
56	Sonar ReasoningPerplexity	Other	14B	▲ 22.7%
57	Simulator v1model.cards	Other	12B	▲ 47.5%
58	Auto Routermodel.cards	Other	0	▲ 0.0%

58 models ranked across the selected window. Traffic figures are illustrative sample data, not measurements of requests served by the gateway.