GPU GUIDE · NVIDIA

Best AI models for the
NVIDIA H200

The NVIDIA H200 has 141GB of VRAM. Below are the top 30 open-source AI models that fit, ranked by composite benchmark score. Each row shows the best quantization that fits your hardware.

VRAM
141GB
Brand
nvidia
Models that fit
30
Generation
Data Center

Top 30 models for the NVIDIA H200

01

Qwen3.8 2.4T A95B

95.0B
77.9
Best fit: Q8_0 · 101.9GBCan I run it? →
02

Qwen3.8 27B

27.0B
69.0
Best fit: fp16 · 55.0GBCan I run it? →
03

GLM-5

230B
54.0
Best fit: Q4_K_M · 140.4GBCan I run it? →
04

Qwen 3.6 Plus

235B
52.5
Best fit: Q4_K_S · 135.5GBCan I run it? →
05

MiMo V2 Omni

120B
49.1
Best fit: Q8_0 · 128.5GBCan I run it? →
06

Qwen 3.6 27B

27.8B
48.3
Best fit: f32 · 112.2GBCan I run it? →
07

Qwen 3 Next 80B A3B

80.0B
45.0
Best fit: Q8_0 · 86.0GBCan I run it? →
08

Qwen3.5-27B

27.0B
44.9
Best fit: fp16 · 55.0GBCan I run it? →
09

Qwen 3.6 35B A3B

36.0B
43.7
Best fit: fp16 · 73.0GBCan I run it? →
10

Muse Glimmer 30B

30.0B
40.6
Best fit: fp16 · 61.0GBCan I run it? →
11

Qwen 3.5 35B A3B

36.0B
37.7
Best fit: fp16 · 73.0GBCan I run it? →
12

Gemma 4 31B

31.3B
37.0
Best fit: f32 · 126.2GBCan I run it? →
13

Devstral 2

24.0B
36.7
Best fit: f32 · 97.0GBCan I run it? →
14

Gemma 4 26B A4B

26.0B
31.9
Best fit: fp16 · 53.0GBCan I run it? →
15

Devstral Small 2

7.0B
31.7
Best fit: f32 · 29.0GBCan I run it? →
16

NVIDIA Nemotron 3 Super 120B A12B BF16

124B
31.1
Best fit: Q8_0 · 132.3GBCan I run it? →
17

Ministral 3 14B

14.0B
26.7
Best fit: f32 · 57.0GBCan I run it? →
18

GPT-OSS 120B

120B
26.0
Best fit: Q8_0 · 128.5GBCan I run it? →
19

gemma 4 12B it

12.0B
25.9
Best fit: f32 · 49.0GBCan I run it? →
20

Ministral 3 8B

8.0B
25.0
Best fit: f32 · 33.0GBCan I run it? →
21

Gemma 4 E2B

5.1B
25.0
Best fit: f32 · 21.4GBCan I run it? →
22

Qwen 3.5 9B

9.7B
24.6
Best fit: f32 · 39.8GBCan I run it? →
23

Qwen3 VL 235B A22B Instruct

236B
24.0
Best fit: Q4_K_S · 135.9GBCan I run it? →
24

Qwen 3.5 4B

4.7B
23.2
Best fit: f32 · 19.8GBCan I run it? →
25

Granite 4.2 8B

8.0B
23.2
Best fit: fp16 · 17.0GBCan I run it? →
26

Mistral Small 4

119B
21.4
Best fit: Q8_0 · 127.4GBCan I run it? →
27

Qwen3 235B A22B Thinking 2507

235B
21.9
Best fit: Q4_K_S · 135.6GBCan I run it? →
28

Devstral 2 2512

125B
21.1
Best fit: Q8_0 · 133.8GBCan I run it? →
29

Devstral Small 2 24B Instruct 2512

24.0B
18.6
Best fit: f32 · 97.0GBCan I run it? →
30

Ministral 3 3B

3.0B
18.3
Best fit: f32 · 13.0GBCan I run it? →
Advertisement

FAQ — running AI on the NVIDIA H200

How many AI models can the NVIDIA H200 run?

With 141GB of VRAM, the NVIDIA H200 can run 30+ open-source models from our database, including Qwen3.8 2.4T A95B, Qwen3.8 27B, GLM-5.

What's the largest LLM I can run on a NVIDIA H200?

The biggest model that fits is approximately 236B. Larger models would need to be quantized further or won't fit at all.

Is 141GB of VRAM enough for local AI?

Yes — 141GB comfortably runs most popular open-source models including 30B-class LLMs at Q4_K_M.