Largest AI Models by Parameter Count

Kimi K3 is the largest listed model at 2.8T parameters. Its size score is 86.

Data sourced from OpenRouter API and manual data collection | Last updated: Nov 3, 2026, 8:01 AM UTC

Ranked by size score

Rank	Model	Parameters	Size score
1	MoonshotAI: Kimi K3	2.8T	86
2	Qwen: Qwen3.8 2.4T A95B	2.4T	85
3	DeepSeek: DeepSeek V4 Pro 0813	1.6T	80
3	Meituan: LongCat 2.0	1.6T	80
5	Mistral: Mistral Large 4	1T	75
5	MoonshotAI: Kimi K2 0711	1T	75
5	MoonshotAI: Kimi K2 0905	1T	75
5	Xiaomi: MiMo-V2.6-Pro	1T	75
9	Thinking Machines: Inkling	975B	75
10	Meta: Muse Spark Open	816B	73
11	Tencent: Hy4 preview	770B	72
12	DeepSeek: DeepSeek V3 0324	685B	71
13	Mistral: Mistral Large 3 2512	675B	71
14	DeepSeek: DeepSeek V3.1	671B	71
14	DeepSeek: R1 0528	671B	71
16	StepFun: Step 5 Preview	600B	69
17	inclusionAI: Ling 3.1 Flash	560B	69
18	NVIDIA: Nemotron 3 Ultra	550B	69
19	Reflection: Beam 1.0 501B A23B	501B	67
20	Qwen: Qwen3 Coder 480B A35B	480B	67
21	MiniMax: MiniMax-01	456B	66
22	Nous: Hermes 4 405B	405B	65
23	Qwen: Qwen3.5 397B A17B	397B	65
24	Xiaomi: MiMo-V2.6-Flash	309B	62
25	Tencent: Hy3	295B	62
26	DeepSeek: DeepSeek V4 Flash 0731	284B	61
27	Thinking Machines: Inkling Small	276B	61
28	Qwen: Qwen3 235B A22B Instruct 2507	235B	59
28	Qwen: Qwen3 235B A22B Thinking 2507	235B	59
28	Qwen: Qwen3 VL 235B A22B Instruct	235B	59
31	MiniMax: MiniMax M2	230B	59
32	StepFun: Step 3.5 Flash	196B	57
32	StepFun: Step 3.7 Flash	196B	57
34	Mistral: Mixtral 8x22B Instruct	176B	56
35	inclusionAI: Ling 3.0 Flash	124B	52
35	inclusionAI: Ling 3.0 Flash VL	124B	52
37	Mistral: Devstral 2 2512	123B	52
38	Qwen: Qwen3.5-122B-A10B	122B	52
39	NVIDIA: Nemotron 3 Super	120B	52
39	NVIDIA: Nemotron 3.5 Super VL	120B	52
41	Mistral: Mistral Small 4	119B	52
42	OpenAI: gpt-oss-120b	117B	52
43	Cohere: Command A	111B	51
44	Meta: Llama 4 Scout	109B	51
45	Z.ai: GLM 4.5V	106B	51
46	Qwen: Qwen3 Coder Next	80B	48
46	Qwen: Qwen3 Next 80B A3B Instruct	80B	48
48	Magnum v4 72B: Magnum v4 72B	72B	46
48	Qwen: Qwen2.5 VL 72B Instruct	72B	46
48	Qwen2.5 72B Instruct: Qwen2.5 72B Instruct	72B	46
51	Meta: Llama 3.1 70B Instruct	70B	46
51	Meta: Llama 3.3 70B Instruct	70B	46
51	Sao10K: Llama 3.1 Euryale 70B v2.2	70B	46
54	TheDrummer: Skyfall 36B V2	36B	39
55	Qwen: Qwen3.5-35B-A3B	35B	39
55	Qwen: Qwen3.6 35B A3B	35B	39
55	Upstage: Solar Mini 4	35B	39
58	Qwen: Qwen3 32B	32.8B	38
59	Qwen2.5 Coder 32B Instruct: Qwen2.5 Coder 32B Instruct	32B	38
60	Google: Gemma 4 31B	31B	37
61	Qwen: Qwen3 30B A3B Instruct 2507	30.5B	37
61	Qwen: Qwen3 Coder 30B A3B Instruct	30.5B	37
63	Meta: Muse Glimmer 30B	30B	37
63	NVIDIA: Nemotron 3 Nano 30B A3B	30B	37
63	NVIDIA: Nemotron 3.5 Lightning	30B	37
63	Qwen: Qwen3 30B A3B	30B	37
63	Qwen: Qwen3 VL 30B A3B Instruct	30B	37
63	Qwen: Qwen3 VL 30B A3B Thinking	30B	37
69	Google: Gemma 2 27B	27B	36
69	Google: Gemma 3 27B	27B	36
69	PrismML: Ternary Bonsai 2 27B	27B	36
69	Qwen: Qwen3.5-27B	27B	36
69	Qwen: Qwen3.6 27B	27B	36
69	Qwen: Qwen3.8 27B	27B	36
75	Google: Gemma 4 26B A4B	25.2B	35
76	Mistral: Mistral Small 3	24B	35
76	Mistral: Mistral Small 3.1 24B	24B	35
76	Mistral: Mistral Small 3.2 24B	24B	35
76	Mistral: Saba	24B	35
76	Mistral: Voxtral Small 24B 2507	24B	35
76	TheDrummer: Cydonia 24B V4.1	24B	35
76	Venice: Uncensored	24B	35
83	OpenAI: gpt-oss-20b	21B	33
83	OpenAI: gpt-oss-safeguard-20b	21B	33
83	Reka Flash 3: Reka Flash 3	21B	33
86	Meta: Llama 4 Maverick	17B	31
87	Qwen: Qwen3 14B	14.8B	29
88	Mistral: Ministral 3 14B 2512	14B	29
89	MythoMax 13B: MythoMax 13B	13B	28
89	ReMM SLERP 13B: ReMM SLERP 13B	13B	28
89	Tencent: Hunyuan A13B Instruct	13B	28
92	Google: Gemma 3 12B	12B	27
92	Meta: Llama Guard 4 12B	12B	27
92	Mistral: Mistral Nemo	12B	27
92	TheDrummer: UnslopNemo 12B	12B	27
96	Moondream: Moondream 3 Preview	9B	24
96	Moondream: Moondream 3.1	9B	24
96	PrismML: Ternary Bonsai 2 9B	9B	24
96	Qwen: Qwen3.5-9B	9B	24
100	AionLabs: Aion-RP 1.0 (8B)	8B	23
100	DeepSeek: DeepSeek V4.1 Flash 1017	8B	23
100	IBM: Granite 4.2 8B	8B	23
100	Meta: Llama 3.1 8B Instruct	8B	23
100	Mistral: Ministral 3 8B 2512	8B	23
100	Qwen: Qwen3 VL 8B Instruct	8B	23
100	Sao10K: Llama 3 8B Lunaris	8B	23
107	ByteDance: UI-TARS 7B	7B	21
107	Qwen: Qwen2.5 7B Instruct	7B	21
109	Google: Gemma 3 4B	4B	15
110	Meta: Llama 3.2 3B Instruct	3B	12
110	Mistral: Ministral 3 3B 2512	3B	12
112	LiquidAI: LFM2.5-2.6B	2.6B	10
113	LiquidAI: LFM2.5-24B-A2B	2.45B	10
114	Meta: Llama 3.2 1B Instruct	1B	0

🤖 Generate website with AI 👆

Join the beta waitlist and get 50% off your first 3 months.

Join waitlist

Frequently asked questions

Which AI model has the most parameters?

Kimi K3 is the largest listed model at 2.8T parameters. Its size score is 86.

How is the size score calculated?

The size score places a published parameter count on a 100-point log scale from 1 billion to 10 trillion. A larger count scores higher.