FreeLLMGPUs › Data center
Best free AI models for the H200
H200: 141 GB of video memory, 4800 GB/s. These are the free models it runs, ranked by Arena votes, with the speed you can expect.
Voted models that fit fully on the GPU
| Model | Arena score | Size | 4-bit file | Tokens/s |
|---|---|---|---|---|
| Gemma 4 31B |
1443 | 31.3 B | 20 GB | 150 |
| Qwen3.8-27B alibaba |
1441 | 27.8 B | 18 GB | 177 |
| Gemma 4 26B A4B google · MoE, 4 B active |
1434 | 25.8 B | 17 GB | 278 |
| Qwen3.5-122B-A10B alibaba · MoE, 10 B active |
1417 | 125.1 B | 77 GB | 224 |
| Qwen3.5-27B alibaba |
1409 | 27.8 B | 18 GB | 177 |
| glm-4.6v zai · MoE, 14.3 B active |
1376 | 107.7 B | 66 GB | 181 |
| gemma-3-27b-it |
1358 | 27.4 B | 18 GB | 174 |
| glm-4.5v zai · MoE, 14.3 B active |
1333 | 107.7 B | 66 GB | 181 |
Bigger models with help from 64 GB of system RAM
Part of the model sits in system memory, which is much slower than video memory.
| Model | Arena score | Size | 4-bit file | Tokens/s |
|---|---|---|---|---|
| MiMo-V2.6-Flash xiaomi · MoE, 17.4 B active |
1455 | 310.8 B | 190 GB | 25 |
| Hy3 tencent · MoE, 24.4 B active |
1438 | 298.8 B | 182 GB | 19 |
| DeepSeek-V4-Flash (High) deepseek · MoE, 20.4 B active |
1432 | 290.9 B | 177 GB | 26 |
| MiMo-V2.5 xiaomi · MoE, 17.4 B active |
1428 | 310.8 B | 190 GB | 25 |
| qwen3-vl-235b-a22b-instruct alibaba · MoE, 22 B active |
1420 | 235.7 B | 144 GB | 87 |
| qwen3-235b-a22b-instruct-2507 alibaba · MoE, 22 B active |
1419 | 235.1 B | 144 GB | 90 |
Popular small models
Official small models with no Arena votes yet, ordered by downloads on Hugging Face (popularity, not quality).
| Model | Downloads/month | Size | 4-bit file | Tokens/s |
|---|---|---|---|---|
| Qwen3-0.6B Qwen |
29.5M | 0.8 B | 1.7 GB | 724 |
| Qwen3-4B Qwen |
11.8M | 4 B | 3.6 GB | 508 |
| Qwen3-8B Qwen |
8.9M | 8.2 B | 6.2 GB | 377 |
| Qwen3-14B Qwen |
8.4M | 14.8 B | 10 GB | 266 |
| Qwen3.5-4B Qwen |
7.8M | 4.7 B | 4.0 GB | 546 |
| Qwen3.5-9B Qwen |
7.5M | 9.7 B | 7.1 GB | 379 |
Estimates at 4-bit (Q4_K_M) with an 8K context, calibrated on 44 real llama.cpp tests (typical error about ±10%). Tokens per second: about 10 is reading speed.
Questions
What is the best free AI model for the H200?
Gemma 4 31B (Arena score 1443) is the highest-voted open model that fits entirely in 141 GB of video memory at 4-bit, writing about 150 tokens per second.
How big a model can the H200 run?
About 218 billion parameters for a dense model at 4-bit with an 8K context, fully on the GPU. Mixture-of-experts models can be bigger, and with 64 GB of system RAM larger models still run, more slowly.
Is the H200 good for local LLMs?
It has 141 GB of memory and 4800 GB/s of bandwidth. Memory decides which models fit; bandwidth decides how fast they write. With 24 GB or more it runs 27–32 B models, the sweet spot of quality for home use.
Similar cards
- Mac M4 Max 128 GB 128 GB · 546 GB/s
- NVIDIA DGX Spark 128 GB 128 GB · 273 GB/s
- Ryzen AI Max+ 395 128 GB 128 GB · 256 GB/s
- RTX PRO 6000 Blackwell 96 GB · 1792 GB/s
- H100 SXM 80 GB · 3350 GB/s
- A100 80 GB 80 GB · 2039 GB/s