Will It Run AI
CalculatorModelsHardwareCompare
Product
  • Calculator
  • Compare
  • Tier List
Browse
  • Models
  • Hardware
  • Docs
About
  • Why It Works
  • What's New
  • Legal Notice
  • Privacy Policy

All estimates are approximations based on mathematical models and public specifications. Actual performance may vary. Do not make purchasing decisions based solely on these estimates.

Data sourced from Hugging Face, Ollama, and official model documentation. Model names and logos are trademarks of their respective owners.

© 2026 Will It Run AI — Fase Consulting Ibiza, S.L. (NIF: B57969656)

Home/Hardware/GPUs/NVIDIA H200 PCIe 141GB
NVIDIA

NVIDIA

NVIDIA H200 PCIe 141GB

Hopper DatacenterDatacenterHopperPCIe 5CUDA
141GB
VRAM
4.8kGB/s
Bandwidth
756TFLOPS
FP16 Compute
1.5kTOPS
INT8 Inference
$30,000 MSRP
VRAM141 GBBandwidth4.8k GB/sCompute756 TFInference1.5k TOPSValue2.52 TF/$k
NVIDIA H200 PCIe 141GBCategory AvgAMD Instinct MI325X 256GB

Specifications

Compute
FP16756 TFLOPS
INT81512 TOPS
ArchitectureHopper
Memory
VRAM141 GB
Bandwidth4800 GB/s
General
FamilyHopper Datacenter
SegmentDatacenter
InterconnectPCIe 5
Compute PlatformCUDA
MSRP$30,000

Architecture

Hopper

Hopper is NVIDIA's datacenter-focused architecture succeeding Ampere. Built on TSMC 4N, it introduces the Transformer Engine with automatic FP8/FP16 mixed-precision training, HBM3/HBM3e memory, and NVLink 4.0 for multi-GPU scaling. The H100 flagship delivers up to 3x the AI training performance of A100.

AI Relevance

The Transformer Engine automatically manages FP8 precision for optimal training speed without accuracy loss. With up to 141 GB HBM3e (H200), Hopper GPUs can hold the largest open-weight models entirely in GPU memory, making them the workhorse of AI datacenters.

Process: TSMC 4NPlatform: CUDATensor Cores: Gen 4Precisions: FP64, FP32, TF32, FP16, BF16, FP8, INT8

Recommendations by Workload

Agentic Coding

B

Mistral Small 4 119B

This model is still usable for agentic-coding, but it is not the most specialized pick. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.

Decode 161.6 tok/s · 50K ctx · llama.cpp
89.6 GB / 141.0 GB VRAM

Chat

B

Mistral Small 4 119B

This model is a direct match for chat. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.

Decode 161.6 tok/s · 13K ctx · llama.cpp
88.4 GB / 141.0 GB VRAM

Coding

B

Mistral Small 4 119B

This model is a direct match for coding. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.

Decode 161.6 tok/s · 25K ctx · llama.cpp
88.6 GB / 141.0 GB VRAM

RAG

B

Command R+ 104B

This model is a direct match for rag. It sits in the middle of the current model mix. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.

Decode 63.6 tok/s · 41K ctx · llama.cpp
110.9 GB / 141.0 GB VRAM

Reasoning

B

Devstral 2 123B Instruct

This model is a direct match for reasoning. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.

Decode 53.7 tok/s · 21K ctx · llama.cpp
109.2 GB / 141.0 GB VRAM

Full Model Compatibility

MistralMixtral 8x22B
B57
141B107.1 GB90 tok/s21K ctx
moe
CohereCommand A 111B
B56
111B100.1 GB60 tok/s23K ctx
dense
MistralMistral Small 4 119B
B56
119B88.6 GB162 tok/s25K ctx
moe
MistralDevstral 2 123B Instruct
B56
123B109.2 GB54 tok/s21K ctx
dense
UnslothQwen3.5 122B A10B
B55
122B93.8 GB63 tok/s24K ctx
dense
AlibabaQwen 2.5 72B
C53
72B70.2 GB92 tok/s32K ctx
dense
AlibabaQwen 2.5 VL 72B
C53
72B70.2 GB92 tok/s32K ctx
dense
MetaLlama 3.3 70B
C53
70B68.6 GB94 tok/s33K ctx
dense
AlibabaQwen3-Coder-Next
C52
80B64.6 GB250 tok/s35K ctx
moe
UnslothQwen3.5 35B A3B
C49
35B41.8 GB189 tok/s54K ctx
dense
Lmstudio-communityLQwen3.5 35B A3B
C48
35B41.8 GB189 tok/s54K ctx
dense
AlibabaQwen 2.5 Coder 32B
C48
32B39.5 GB207 tok/s57K ctx
dense
UnslothQwen3.5 27B
C48
27B35.7 GB245 tok/s63K ctx
dense
Unslothgemma 3 27b it
C48
27B35.7 GB245 tok/s63K ctx
dense
AlibabaQwen3-Coder 30B A3B Instruct
C47
30.5B34.4 GB561 tok/s66K ctx
moe
AlibabaQwen3-VL 30B A3B Instruct
C47
30B34.1 GB580 tok/s66K ctx
moe
MistralDevstral Small 2 24B Instruct
C47
24B33.4 GB275 tok/s68K ctx
dense
MistralDevstral Small 1.1
C47
24B33.4 GB275 tok/s68K ctx
dense
MistralCodestral 2 25.08
C47
22B31.9 GB300 tok/s71K ctx
dense
UnslothQwen3.5 9B
C46
9B21.9 GB734 tok/s103K ctx
dense
HauhauCSHQwen3.5 9B Uncensored HauhauCS Aggressive
C46
9B21.9 GB734 tok/s103K ctx
dense
Lmstudio-communityLQwen3.5 9B
C45
9B21.9 GB734 tok/s103K ctx
dense
BartowskiBMeta Llama 3.1 8B Instruct
C45
8B21.1 GB826 tok/s107K ctx
dense
XtunerXllava llama 3 8b v1 1
C45
8B21.1 GB826 tok/s107K ctx
dense
UnslothDeepSeek R1 0528 Qwen3 8B
C45
8B21.1 GB826 tok/s107K ctx
dense
MaziyarPanahiMMeta Llama 3 8B Instruct
C45
8B21.1 GB826 tok/s107K ctx
dense
TheBlokeTLlama 2 7B Chat
C45
7B20.4 GB944 tok/s111K ctx
dense
TheBlokeTMistral 7B Instruct v0.2
C45
7B20.4 GB944 tok/s111K ctx
dense
MaziyarPanahiMMistral 7B Instruct v0.3
C45
7B20.4 GB944 tok/s111K ctx
dense
UnslothQwen3.5 4B
C45
4B18.2 GB1653 tok/s124K ctx
dense
BartowskiBLlama 3.2 3B Instruct
C45
3B18.0 GB1904 tok/s126K ctx
dense
BartowskiBgemma 2 2b it
C45
2B17.4 GB2582 tok/s129K ctx
dense
Lmstudio-communityLgemma 3 4b it
C45
4B18.2 GB1653 tok/s124K ctx
dense
QwenQwen2.5 3B Instruct
C45
3B17.6 GB2203 tok/s128K ctx
dense
Googlegemma 2b
C45
2B17.0 GB3305 tok/s133K ctx
dense
Hugging-quantsHLlama 3.2 1B Instruct Q8 0
C45
1B16.6 GB4234 tok/s136K ctx
dense
TheDrummerTGemmasutra Mini 2B v1
C45
2B17.0 GB3305 tok/s133K ctx
dense
QwenQwen2.5 1.5B Instruct
C45
1.5B16.7 GB4032 tok/s135K ctx
dense
TheBlokeTTinyLlama 1.1B Chat v1.0
C45
1.1B16.5 GB4032 tok/s137K ctx
dense
Ggml-orgGSmolVLM 500M Instruct
C45
0.5B16.2 GB4234 tok/s139K ctx
dense
Ggml-orgGembeddinggemma 300M
C45
0.3B16.0 GB4234 tok/s141K ctx
dense
AlibabaQwen 3 235B A22B
C43
235B161.8 GB67 tok/s14K ctx
moe
DeepSeekDeepSeek R1 671B
F0
671B430.1 GB29 tok/s5K ctx
moe
Z.aiGLM-5
F0
744B475.1 GB26 tok/s5K ctx
moe
Moonshot AIKimi K2.5
F0
1000B630.0 GB20 tok/s4K ctx
moe
MistralMistral Large 3
F0
675B433.2 GB28 tok/s5K ctx
+1moe
AlibabaQwen3-Coder 480B A35B Instruct
F0
480B313.3 GB38 tok/s7K ctx
moe
DeepSeekDeepSeek V3 671B
F0
671B430.1 GB29 tok/s5K ctx
moe
UnslothQwen3.5 397B A17B
F0
397B319.2 GB17 tok/s7K ctx
dense
MetaLlama 4 Maverick 17B 128E
F0
400B261.7 GB50 tok/s9K ctx
moe

Just out of reach

Models you could run with an upgrade

High-quality models that need a bit more memory

DeepSeekDeepSeek R1 671B
671BTier 5Needs ~435.9 GB
Z.aiGLM-5
744BTier 5Needs ~481.3 GB
Moonshot AIKimi K2.5
1000BTier 5Needs ~635.0 GB
MistralMistral Large 3
675BTier 5Needs ~439.6 GB
AlibabaQwen3-Coder 480B A35B Instruct
480BTier 5Needs ~318.7 GB

Upgrade paths

Upgrade from NVIDIA H200 PCIe 141GB

See what you unlock with more powerful hardware

Upgrade options

Upgrade options

AMDAMD Instinct MI325X 256GBNext step up
256 GB VRAM (+115)6000 GB/s (+1200)
A
Unlocks Llama 4 Maverick 17B 128E, K EXAONE 236B A23B, Baichuan M3 235B+1 more · +7% faster avg

 

AMDAMD Instinct MI350X 288GBBest value
288 GB VRAM (+147)8000 GB/s (+3200)
A
Unlocks Qwen3-Coder 480B A35B Instruct, Llama 4 Maverick 17B 128E, K EXAONE 236B A23B+2 more · +43% faster avg

~$8,000 MSRP

Compare this GPU