Your rig

H100 SXM · 80 GB

128 GB system RAM · Hopper · 3350 GB/s · 78.4 GB usable VRAM
Edit hardware Export as JSON

Context length

KV cache at f16, 32K tokens

Quality floor
Q8_0 — near-lossless Q6_K Q5_K_M Q4_K_M — recommended Q3_K_M Q2_K — damaged
KV cache type
System RAM (GB)
Runtime
Ollamallama.cppLM StudiovLLM

4 of 5 runtimes drive this device.

Allow CPU offload Include partial-offload fits

Offloaded layers run at 60 GB/s host bandwidth — expect single-digit tokens per second.

Which one should I actually run?

Best models for your H100 SXM

Updated 8 Oct 2026 1 new this month

What do you want to do? Each row is the best-ranked model on that use case's shortlist that fits at Q2_K and 32K. The ordering is editorial; the sizes and speeds are computed. How the shortlists work.

RecommendationModelWhyQuant · totalTok/s est.
Best overall Qwen3.8 27B
Alibaba · Apache 2.0
Best capability per gigabyte: a current-generation dense 27B with vision and 262K context. Q2_K · 13.4 GB 187 Runs great
Best for coding Ling 3.0 Flash 124B-A5B
inclusionAI · MIT
SWE-bench Pro 56.6 claimed at 5B active — the 128 GB-class coding pick if you cannot fit DeepSeek V4. Q2_K · 49.2 GB 392 Runs great
Best reasoning Ling 3.0 Flash 124B-A5B
inclusionAI · MIT
AIME 93 claimed at 5B active; hybrid attention keeps long reasoning traces cheap. Q2_K · 49.2 GB 392 Runs great
Best for writing Llama 3.3 70B Instruct
Meta · Llama 3.3 Community
The creative-writing favourite: consistent voice, takes direction, handles fiction framing. Q2_K · 38.1 GB 74 Runs great
Best vision Qwen3.8 27B
Alibaba · Apache 2.0
Image and video in, OSWorld 84 — the strongest local vision-language model at 24 GB. Q2_K · 13.4 GB 187 Runs great
Best for agents Ling 3.0 Flash 124B-A5B
inclusionAI · MIT
Agentic workflows are what Ling 3.0 was trained for; MIT, 5B active. Q2_K · 49.2 GB 392 Runs great
Best translation Qwen3.8 27B
Alibaba · Apache 2.0
119 languages, and Qwen translations read as idiomatic rather than literal. Q2_K · 13.4 GB 187 Runs great
Fastest good model Qwen3.6 35B-A3B
Alibaba · Apache 2.0
3.3B active per token: the fastest model that still competes with dense 27Bs. Q2_K · 15.2 GB 606 Runs great
Best long context Qwen3.8 27B
Alibaba · Apache 2.0
262K native, and only 16 of 64 blocks keep a KV cache — long context is cheap here. Q2_K · 19.4 GB
at 128K context
187 Runs great

Everything that fits

Runs great

66 of 80 · fits with headroom for a longer prompt
ModelParamsQuantWeights +KV 32KTotal VRAM headroom Tok/sRuntime
Qwen3.5 122B-A10B 125B MoE / 10B act Q2_K 48.7 GB 0.75 GB 50.1 GB
28.3 GB free
200 Ollama Run
Nemotron 3 Super 120B-A12B 124B MoE / 12B act Q2_K 48.4 GB 0.25 GB 49.2 GB
29.2 GB free
167 llama.cpp Run
Ling 3.0 Flash 124B-A5B 124B MoE / 5.1B act Q2_K 48.4 GB 0.25 GB 49.2 GB
29.2 GB free
392 llama.cpp Run
Mistral Small 4 119B-A6B 119B MoE / 6.5B act Q2_K 46.4 GB 0.70 GB 47.7 GB
30.7 GB free
308 llama.cpp Run
gpt-oss 120B 117B MoE / 5.1B act MXFP4 60.5 GB 1.13 GB 62.2 GB
16.2 GB free
295 Ollama Run
Llama 4 Scout 109B-A17B 109B MoE / 17B act Q2_K 42.5 GB 2.63 GB 45.7 GB
32.7 GB free
118 Ollama Run
Llama 3.3 70B Instruct 70.6B Q2_K 27.5 GB 10.00 GB 38.1 GB
40.3 GB free
74 Ollama Run
DeepSeek-R1-Distill-Llama 70B 70.6B Q2_K 27.5 GB 10.00 GB 38.1 GB
40.3 GB free
74 Ollama Run
Qwen3.6 35B-A3B 35.9B MoE / 3.3B act Q2_K 14.0 GB 0.63 GB 15.2 GB
63.2 GB free
606 Ollama Run
Ornith 1.5 35B-A3B 35.9B MoE / 3.3B act Q2_K 14.0 GB 0.63 GB 15.2 GB
63.2 GB free
606 Ollama Run
LLM-jp 4 33B Thinking 33.2B Q2_K 12.9 GB 8.00 GB 21.5 GB
56.9 GB free
157 llama.cpp Run
Qwen3 32B 32.8B Q2_K 12.8 GB 8.00 GB 21.4 GB
57.0 GB free
159 Ollama Run
DeepSeek-R1-Distill-Qwen 32B 32.8B Q2_K 12.8 GB 8.00 GB 21.4 GB
57.0 GB free
159 Ollama Run
Qwen2.5-Coder 32B 32.8B Q2_K 12.8 GB 8.00 GB 21.4 GB
57.0 GB free
159 Ollama Run
Olmo 3.1 32B Instruct 32.2B Q2_K 12.6 GB 2.75 GB 15.9 GB
62.5 GB free
161 Ollama Run
Nemotron 3.5 Lightning 30B-A3B 31.6B MoE / 3.2B act Q2_K 12.3 GB 0.19 GB 13.1 GB
65.3 GB free
625 Ollama Run
Gemma 4 31B 31.3B Q2_K 12.2 GB 5.78 GB 18.6 GB
59.8 GB free
166 Ollama Run
GLM-4.7-Flash 30B-A3B 31.2B MoE / 3B act Q2_K 12.2 GB 1.65 GB 14.4 GB
64.0 GB free
667 Ollama Run
Qwen3 30B-A3B 30.5B MoE / 3.3B act Q2_K 11.9 GB 3.00 GB 15.5 GB
62.9 GB free
606 Ollama Run
Qwen3 Coder 30B-A3B 30.5B MoE / 3.3B act Q2_K 11.9 GB 3.00 GB 15.5 GB
62.9 GB free
606 Ollama Run
Muse Glimmer 30B 29.8B Q2_K 11.6 GB 0.48 GB 12.7 GB
65.7 GB free
174 Ollama Run
Granite 4.1 30B 28.9B Q2_K 11.3 GB 8.00 GB 19.9 GB
58.5 GB free
180 Ollama Run
Qwen3.8 27B 27.8B Q2_K 10.8 GB 2.00 GB 13.4 GB
65.0 GB free
187 Ollama Run
Qwen3.6 27B 27.8B Q2_K 10.8 GB 2.00 GB 13.4 GB
65.0 GB free
187 Ollama Run
Gemma 3 27B 27.4B Q2_K 10.7 GB 2.91 GB 14.2 GB
64.2 GB free
190 Ollama Run
Gemma 4 26B-A4B 26.5B MoE / 3.8B act Q2_K 10.3 GB 1.45 GB 12.4 GB
66.0 GB free
526 Ollama Run
Devstral Small 2 24B 24B Q2_K 9.4 GB 5.00 GB 15.0 GB
63.4 GB free
217 llama.cpp Run
Mistral Small 3.2 24B 23.6B Q2_K 9.2 GB 5.00 GB 14.8 GB
63.6 GB free
220 Ollama Run
gpt-oss 20B 20.9B MoE / 3.6B act MXFP4 10.8 GB 0.75 GB 12.2 GB
66.2 GB free
418 Ollama Run
Qwen3 14B 14.8B Q2_K 5.8 GB 5.00 GB 11.4 GB
67.0 GB free
351 Ollama Run
DeepSeek-R1-Distill-Qwen 14B 14.8B Q2_K 5.8 GB 6.00 GB 12.4 GB
66.0 GB free
351 Ollama Run
Qwen2.5-Coder 14B 14.8B Q2_K 5.8 GB 6.00 GB 12.4 GB
66.0 GB free
351 Ollama Run
Phi-4 14B 14.7B Q2_K 5.7 GB 3.13 GB 9.5 GB
68.9 GB free
354 Ollama Run
Ministral 3 14B 13.9B Q2_K 5.4 GB 5.00 GB 11.0 GB
67.4 GB free
374 Ollama Run
Gemma 3 12B 12.2B Q2_K 4.8 GB 2.31 GB 7.7 GB
70.7 GB free
426 Ollama Run
Mistral NeMo 12B 12.2B Q2_K 4.8 GB 5.00 GB 10.4 GB
68.0 GB free
426 Ollama Run
Gemma 4 12B 12B Q2_K 4.7 GB 2.31 GB 7.6 GB
70.8 GB free
433 Ollama Run
Qwen3.5 9B 9.65B Q2_K 3.8 GB 1.00 GB 5.4 GB
73.0 GB free
539 Ollama Run
Ornith 1.5 9B 9.41B Q2_K 3.7 GB 1.00 GB 5.3 GB
73.1 GB free
553 Ollama Run
Ministral 3 8B 8.92B Q2_K 3.5 GB 4.25 GB 8.3 GB
70.1 GB free
583 Ollama Run
Granite 4.1 8B 8.79B Q2_K 3.4 GB 5.00 GB 9.0 GB
69.4 GB free
592 Ollama Run
LFM2.5 8B-A1B 8.47B MoE / 1.5B act Q2_K 3.3 GB 0.38 GB 4.3 GB
74.1 GB free
1333 Ollama Run
Fara 7B 8.29B Q2_K 3.2 GB 1.75 GB 5.6 GB
72.8 GB free
627 llama.cpp Run
Qwen3 8B 8.19B Q2_K 3.2 GB 4.50 GB 8.3 GB
70.1 GB free
635 Ollama Run
Llama 3.1 8B Instruct 8.03B Q2_K 3.1 GB 4.00 GB 7.7 GB
70.7 GB free
648 Ollama Run
Gemma 4 E4B 8.0B Q2_K 3.1 GB 0.47 GB 4.2 GB
74.2 GB free
650 Ollama Run
Ling 3.0 Tiny 7.9B-A1.3B 7.9B MoE / 1.3B act Q2_K 3.1 GB 0.21 GB 3.9 GB
74.5 GB free
1538 llama.cpp Run
DeepSeek-R1-Distill-Qwen 7B 7.62B Q2_K 3.0 GB 1.75 GB 5.3 GB
73.1 GB free
682 Ollama Run
Qwen2.5-Coder 7B 7.62B Q2_K 3.0 GB 1.75 GB 5.3 GB
73.1 GB free
682 Ollama Run
Olmo 3 7B Instruct 7.3B Q2_K 2.8 GB 5.50 GB 8.9 GB
69.5 GB free
712 Ollama Run
Mistral 7B Instruct v0.3 7.25B Q2_K 2.8 GB 4.00 GB 7.4 GB
71.0 GB free
717 Ollama Run
Gemma 4 E2B 5.1B Q2_K 2.0 GB 0.23 GB 2.8 GB
75.6 GB free
1020 Ollama Run
Qwen3.5 4B 4.66B Q2_K 1.8 GB 1.00 GB 3.4 GB
75.0 GB free
1116 Ollama Run
Gemma 3 4B 4.3B Q2_K 1.7 GB 0.86 GB 3.1 GB
75.3 GB free
1209 Ollama Run
Qwen3 4B 4.02B Q2_K 1.6 GB 4.50 GB 6.7 GB
71.7 GB free
1294 Ollama Run
Ministral 3 3B 3.85B Q2_K 1.5 GB 3.25 GB 5.4 GB
73.0 GB free
1351 Ollama Run
Phi-4-mini 3.8B 3.84B Q2_K 1.5 GB 4.00 GB 6.1 GB
72.3 GB free
1354 Ollama Run
Granite 4.1 3B 3.4B Q2_K 1.3 GB 2.50 GB 4.4 GB
74.0 GB free
1529 Ollama Run
Llama 3.2 3B Instruct 3.21B Q2_K 1.3 GB 3.50 GB 5.4 GB
73.0 GB free
1620 Ollama Run
LFM2.5 2.6B 2.7B Q2_K 1.1 GB 0.50 GB 2.2 GB
76.2 GB free
1926 llama.cpp Run
Qwen3.5 2B 2.27B Q2_K 0.9 GB 0.38 GB 1.9 GB
76.5 GB free
2291 Ollama Run
Qwen3 1.7B 1.72B Q2_K 0.7 GB 3.50 GB 4.8 GB
73.6 GB free
3023 Ollama Run
Llama 3.2 1B Instruct 1.24B Q2_K 0.5 GB 1.00 GB 2.1 GB
76.3 GB free
4194 Ollama Run
Gemma 3 1B 1.0B Q2_K 0.4 GB 0.14 GB 1.1 GB
77.3 GB free
5200 Ollama Run
Qwen3.5 0.8B 0.87B Q2_K 0.3 GB 0.38 GB 1.3 GB
77.1 GB free
5977 Ollama Run
Qwen3 0.6B 0.6B Q2_K 0.2 GB 3.50 GB 4.3 GB
74.1 GB free
8667 Ollama Run

Runs — tight

1 model · fits at 32K, drop context or quant to go longer
ModelParamsQuantWeights +KV 32KTotal VRAM headroom Tok/sRuntime
Qwen3.8-Flash-Next 180B-A6B 180B MoE / 6B act Q2_K 70.2 GB 0.75 GB 71.5 GB
6.9 GB free
333 llama.cpp Run

Won't fit in VRAM

13 models · some are possible with CPU offload
ModelParamsQuantWeights +KV 32KTotal Over budget Tok/sRuntime
Qwen3.5 397B-A17B 403B MoE / 17B act Q2_K 157.2 GB 0.94 GB 158.7 GB +80.3 GB ~4.1 llama.cpp Why
Ornith 1.5 397B-A17B 397B MoE / 17B act Q2_K 154.8 GB 0.94 GB 156.4 GB +78.0 GB ~4.1 llama.cpp Why
GLM-5.3-Flash 320B-A18B 321B MoE / 18B act Q2_K 125.2 GB 0.34 GB 126.1 GB +47.7 GB ~5.1 llama.cpp Why
DeepSeek V4 Flash 284B-A13B 284B MoE / 13B act Q2_K 110.8 GB 1.51 GB 112.9 GB +34.5 GB ~8.5 llama.cpp Why
Qwen3 235B-A22B 235B MoE / 22B act Q2_K 91.6 GB 5.88 GB 98.1 GB +19.7 GB ~7.1 llama.cpp Why
Kimi K3 2.8T-A104B 2780B MoE / 104B act Q2_K 1084.2 GB 0.84 GB 1085.6 GB +1007.2 GB ~0.4 llama.cpp Why
Qwen3.8 2.4T-A95B 2446B MoE / 95B act Q2_K 953.9 GB 2.88 GB 957.4 GB +879.0 GB ~0.4 llama.cpp Why
DeepSeek V4 Pro 1.6T-A49B 1650B MoE / 49B act Q2_K 643.5 GB 2.14 GB 646.2 GB +567.8 GB ~0.8 llama.cpp Why
Kimi K2.6 1T-A32B 1027B MoE / 32B act Q2_K 400.5 GB 2.14 GB 403.3 GB +324.9 GB ~1.4 llama.cpp Why
GLM-5.2 744B-A40B 753B MoE / 40B act Q2_K 293.7 GB 2.74 GB 297.0 GB +218.6 GB ~1.2 llama.cpp Why
GLM-5.3 744B-A40B 753B MoE / 40B act Q2_K 293.7 GB 2.74 GB 297.0 GB +218.6 GB ~1.2 llama.cpp Why
DeepSeek-R1 671B 671B MoE / 37B act Q2_K 261.7 GB 2.14 GB 264.4 GB +186.0 GB ~1.4 llama.cpp Why
DeepSeek V4.1 Flash 552B-A16B New 552B MoE / 16B act Q2_K 215.3 GB 0.15 GB 216.0 GB +137.6 GB ~3.5 vLLM Why

Totals = quantised weights + f16 KV cache at 32K tokens + 0.6 GB runtime overhead, against 78.4 GB usable VRAM (80 GB less display and driver reserve). Tokens per second are estimated from 3350 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.


Same memory, different speed

These devices have the same 78.4 GB usable, so they run exactly the models above. Generation speed scales with memory bandwidth.

DeviceBandwidthSpeed vs yours
A100 80 GB
Ampere · 2021
2039 GB/s 0.61×
H100 PCIe
Hopper · 2022
2000 GB/s 0.60×