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/Gaudi 3 128GB
Intel

Intel

Gaudi 3 128GB

Data CenterGaudiPCIe 5oneAPI
128GB
VRAM
3.7kGB/s
Bandwidth
900TFLOPS
FP16 Compute
1.8kTOPS
INT8 Inference
VRAM128 GBBandwidth3.7k GB/sCompute900 TFInference1.8k TOPS
Gaudi 3 128GBCategory AvgNVIDIA H200 141GB

Specifications

Compute
FP16900 TFLOPS
INT81835 TOPS
ArchitectureGaudi
Memory
VRAM128 GB
Bandwidth3700 GB/s
General
FamilyData Center
SegmentData Center
InterconnectPCIe 5
Compute PlatformONEAPI

Architecture

Gaudi

Gaudi is Intel's purpose-built AI training and inference accelerator (acquired from Habana Labs). Gaudi 3 features 128 GB HBM2e and a dedicated Matrix Math Engine designed specifically for transformer workloads.

AI Relevance

Purpose-built for AI with integrated networking (24x 200GbE) for multi-node scaling. Gaudi 3 targets direct competition with NVIDIA H100 for transformer training and inference, with competitive TCO claims.

Process: TSMC 5nmPlatform: ONEAPIPrecisions: 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 103.8 tok/s · 46K ctx · llama.cpp
88.3 GB / 128.0 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 103.8 tok/s · 12K ctx · llama.cpp
87.1 GB / 128.0 GB VRAM

Coding

C

Qwen3-Coder-Next

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, ollama, lm-studio.

Decode 160.8 tok/s · 32K ctx · llama.cpp
63.3 GB / 128.0 GB VRAM

RAG

C

Command R 35B

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 121.3 tok/s · 89K ctx · llama.cpp
46.0 GB / 128.0 GB VRAM

Reasoning

B

Mistral Small 4 119B

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 103.8 tok/s · 23K ctx · llama.cpp
87.3 GB / 128.0 GB VRAM

Full Model Compatibility

MistralMistral Small 4 119B
B57
119B87.3 GB104 tok/s23K ctx
moe
UnslothQwen3.5 122B A10B
C55
122B92.5 GB40 tok/s22K ctx
dense
CohereCommand A 111B
C54
111B98.8 GB38 tok/s21K ctx
dense
AlibabaQwen3-Coder-Next
C53
80B63.3 GB161 tok/s32K ctx
moe
MistralMixtral 8x22B
C53
141B105.8 GB58 tok/s19K ctx
moe
AlibabaQwen 2.5 72B
C52
72B68.9 GB59 tok/s30K ctx
dense
AlibabaQwen 2.5 VL 72B
C52
72B68.9 GB59 tok/s30K ctx
dense
MetaLlama 3.3 70B
C52
70B67.3 GB61 tok/s30K ctx
dense
MistralDevstral 2 123B Instruct
C51
123B107.9 GB35 tok/s19K ctx
dense
UnslothQwen3.5 35B A3B
C49
35B40.5 GB121 tok/s51K ctx
dense
Lmstudio-communityLQwen3.5 35B A3B
C49
35B40.5 GB121 tok/s51K ctx
dense
AlibabaQwen 2.5 Coder 32B
C49
32B38.2 GB133 tok/s54K ctx
dense
UnslothQwen3.5 27B
C48
27B34.4 GB157 tok/s60K ctx
dense
Unslothgemma 3 27b it
C48
27B34.4 GB157 tok/s60K ctx
dense
AlibabaQwen3-Coder 30B A3B Instruct
C48
30.5B33.1 GB360 tok/s62K ctx
moe
AlibabaQwen3-VL 30B A3B Instruct
C48
30B32.8 GB372 tok/s62K ctx
moe
MistralDevstral Small 2 24B Instruct
C48
24B32.1 GB177 tok/s64K ctx
dense
MistralDevstral Small 1.1
C47
24B32.1 GB177 tok/s64K ctx
dense
MistralCodestral 2 25.08
C47
22B30.6 GB193 tok/s67K ctx
dense
UnslothQwen3.5 9B
C46
9B20.6 GB472 tok/s99K ctx
dense
HauhauCSHQwen3.5 9B Uncensored HauhauCS Aggressive
C46
9B20.6 GB472 tok/s99K ctx
dense
Lmstudio-communityLQwen3.5 9B
C46
9B20.6 GB472 tok/s99K ctx
dense
BartowskiBMeta Llama 3.1 8B Instruct
C46
8B19.8 GB531 tok/s103K ctx
dense
XtunerXllava llama 3 8b v1 1
C45
8B19.8 GB531 tok/s103K ctx
dense
UnslothDeepSeek R1 0528 Qwen3 8B
C45
8B19.8 GB531 tok/s103K ctx
dense
MaziyarPanahiMMeta Llama 3 8B Instruct
C45
8B19.8 GB531 tok/s103K ctx
dense
TheBlokeTLlama 2 7B Chat
C45
7B19.1 GB607 tok/s107K ctx
dense
TheBlokeTMistral 7B Instruct v0.2
C45
7B19.1 GB607 tok/s107K ctx
dense
MaziyarPanahiMMistral 7B Instruct v0.3
C45
7B19.1 GB607 tok/s107K ctx
dense
UnslothQwen3.5 4B
C45
4B16.9 GB1062 tok/s121K ctx
dense
BartowskiBLlama 3.2 3B Instruct
C45
3B16.7 GB1223 tok/s123K ctx
dense
BartowskiBgemma 2 2b it
C45
2B16.1 GB1658 tok/s127K ctx
dense
Lmstudio-communityLgemma 3 4b it
C45
4B16.9 GB1062 tok/s121K ctx
dense
QwenQwen2.5 3B Instruct
C45
3B16.3 GB1415 tok/s125K ctx
dense
Googlegemma 2b
C45
2B15.7 GB2123 tok/s130K ctx
dense
TheDrummerTGemmasutra Mini 2B v1
C45
2B15.7 GB2123 tok/s130K ctx
dense
Hugging-quantsHLlama 3.2 1B Instruct Q8 0
C45
1B15.3 GB2720 tok/s134K ctx
dense
QwenQwen2.5 1.5B Instruct
C45
1.5B15.4 GB2590 tok/s133K ctx
dense
TheBlokeTTinyLlama 1.1B Chat v1.0
C45
1.1B15.2 GB2590 tok/s135K ctx
dense
Ggml-orgGSmolVLM 500M Instruct
C45
0.5B14.9 GB2720 tok/s137K ctx
dense
Ggml-orgGembeddinggemma 300M
C45
0.3B14.7 GB2720 tok/s139K ctx
dense
DeepSeekDeepSeek R1 671B
F0
671B428.8 GB18 tok/s5K ctx
moe
Z.aiGLM-5
F0
744B473.8 GB17 tok/s4K ctx
moe
Moonshot AIKimi K2.5
F0
1000B628.7 GB13 tok/s4K ctx
moe
MistralMistral Large 3
F0
675B431.9 GB18 tok/s5K ctx
+1moe
AlibabaQwen3-Coder 480B A35B Instruct
F0
480B312.0 GB25 tok/s7K ctx
moe
DeepSeekDeepSeek V3 671B
F0
671B428.8 GB18 tok/s5K ctx
moe
AlibabaQwen 3 235B A22B
F0
235B160.5 GB48 tok/s13K ctx
moe
UnslothQwen3.5 397B A17B
F0
397B317.9 GB11 tok/s6K ctx
dense
MetaLlama 4 Maverick 17B 128E
F0
400B260.4 GB32 tok/s8K 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 ~434.6 GB
Z.aiGLM-5
744BTier 5Needs ~480.0 GB
Moonshot AIKimi K2.5
1000BTier 5Needs ~633.7 GB
MistralMistral Large 3
675BTier 5Needs ~438.3 GB
AlibabaQwen3-Coder 480B A35B Instruct
480BTier 5Needs ~317.4 GB

Upgrade paths

Upgrade from Gaudi 3 128GB

See what you unlock with more powerful hardware

Upgrade options

Upgrade options

NVIDIANVIDIA H200 141GBNext step up
141 GB VRAM (+13)4800 GB/s (+1100)
A
Unlocks Qwen 3 235B A22B, DeepSeek Coder V2 236B, DeepSeek V2.5 236B+54% faster avg

~$30,000 MSRP

AMDAMD Instinct MI325X 256GBBiggest leap
256 GB VRAM (+128)6000 GB/s (+2300)
A
Unlocks Qwen 3 235B A22B, Llama 4 Maverick 17B 128E, DeepSeek Coder V2 236B+4 more · +66% faster avg

 

AMDAMD Instinct MI350X 288GBBest value
288 GB VRAM (+160)8000 GB/s (+4300)
A
Unlocks Qwen3-Coder 480B A35B Instruct, Qwen 3 235B A22B, Llama 4 Maverick 17B 128E+5 more · +120% faster avg

~$8,000 MSRP

Compare this GPU