AI Compatibility Tier List
NVIDIA GB200 192GB
374 AI models ranked by how well they run on your 192 GB VRAM hardware. Best match: Devstral 2 123B Instruct (score: 96).
374
Models tested
338
Can run
192 GB
VRAM
97
Best tok/s
Tier Distribution
S 27
A 55
B 65
C 191
F 36
S: 27A: 55B: 65C: 191F: 36
S
S Tier -- Perfect Fit (27 models)
These models run flawlessly with headroom to spare.
Devstral 2 123B Instruct123B100.5 GB97 tok/s256K ctx
dense
DeepSeek V4 Flash284B179.4 GB145 tok/s169K ctx
moe
Qwen 3.5 122B A10B122B97.0 GB270 tok/s131K ctx
moe
Mistral Small 4 119B119B98.1 GB293 tok/s256K ctx
moe
GPT-OSS 120B117B96.4 GB102 tok/s131K ctx
dense
Pixtral Large 124B124B101.1 GB97 tok/s131K ctx
dense
Command A 111B111B91.7 GB108 tok/s262K ctx
dense
Qwen 3 235B A22B235B166.3 GB137 tok/s131K ctx
moe
MiniMax M2.7
S90230B164.2 GB156 tok/s134K ctx
moe
Qwen 2.5 VL 72B72B68.9 GB166 tok/s33K ctx
dense
Qwen3-Coder-Next80B70.4 GB454 tok/s256K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B40.2 GB1016 tok/s256K ctx
moe
Qwen 3.6 35B A3B35B45.6 GB854 tok/s262K ctx
moe
Leanstral 119B A6B119B101.5 GB269 tok/s181K ctx
moe
Qwen 3.5 27B27B39.7 GB378 tok/s131K ctx
dense
Qwen3-VL 30B A3B Instruct30B39.9 GB1051 tok/s256K ctx
moe
Qwen 3.6 27B27B37.5 GB275 tok/s262K ctx
dense
Qwen 3.5 35B A3B35B42.9 GB929 tok/s131K ctx
moe
Qwen 3 32B32B43.5 GB374 tok/s131K ctx
dense
Magistral Small 250724B37.2 GB336 tok/s131K ctx
dense
Devstral Small 2 24B Instruct24B37.2 GB336 tok/s256K ctx
dense
Qwen 3 30B A3B30.5B40.2 GB1016 tok/s131K ctx
moe
Qwen 3.5 9B9B27.8 GB126 tok/s131K ctx
dense
Nemotron 3 Nano 30B30B40.8 GB395 tok/s131K ctx
dense
Qwen 3 14B14B31.1 GB196 tok/s131K ctx
dense
Devstral Small 1.124B37.2 GB336 tok/s131K ctx
dense
Phi-4-reasoning-plus 14B14.7B32.1 GB206 tok/s33K ctx
dense
A
A Tier -- Great Fit (55 models)
Excellent performance with comfortable memory margins.
Qwen 3 8B8B27.2 GB112 tok/s131K ctx
dense
Gemma 4 31B30.7B53.5 GB234 tok/s167K ctx
dense
GPT-OSS 20B21B35.4 GB1290 tok/s128K ctx
moe
Nemotron Cascade 2 30B A3B30B41.3 GB1039 tok/s262K ctx
moe
Llama 3.3 70B70B67.7 GB171 tok/s128K ctx
dense
Qwen 3.5 4B4B24.7 GB56 tok/s131K ctx
dense
EXAONE 4.0 32B32B43.5 GB372 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B39.1 GB1091 tok/s256K ctx
moe
Qwen 2.5 72B72B68.9 GB166 tok/s131K ctx
dense
Llama 3.1 70B70B67.7 GB171 tok/s128K ctx
dense
Ministral 3 14B14B31.1 GB196 tok/s262K ctx
multimodal
Qwen 2.5 32B32B43.5 GB372 tok/s131K ctx
dense
Codestral 2 25.0822B36.0 GB308 tok/s256K ctx
dense
Mistral Small 3.2 24B24B37.2 GB336 tok/s131K ctx
vision
Nemotron Nano 8B8B26.9 GB112 tok/s131K ctx
dense
Phi-4 Mini Reasoning 4B3.8B23.9 GB53 tok/s131K ctx
dense
Kimi Linear 48B A3B48B50.3 GB230 tok/s1049K ctx
linear
Gemma 3 27B27B47.8 GB266 tok/s131K ctx
dense
Llama 4 Scout 17B 16E109B89.5 GB257 tok/s576K ctx
moe
DeepSeek Coder V2 236B236B222.7 GB84 tok/s8K ctx
moe
LFM2 24B24B37.2 GB336 tok/s131K ctx
dense
OLMo 2 32B32B43.5 GB372 tok/s4K ctx
dense
Granite 4.1 30B30B42.3 GB395 tok/s131K ctx
dense
CogVLM2 19B19B34.1 GB266 tok/s8K ctx
dense
Mistral Small 24B24B37.2 GB336 tok/s33K ctx
dense
DeepSeek V2.5 236B236B222.7 GB84 tok/s8K ctx
moe
Mistral Small 3.1 24B24B37.2 GB336 tok/s131K ctx
dense
InternVL2 8B8B26.9 GB112 tok/s8K ctx
dense
Phi-4 14B14B31.7 GB196 tok/s16K ctx
dense
Qwen 2.5 14B14B31.6 GB196 tok/s131K ctx
dense
DeepSeek R1 Distill 70B70B67.7 GB171 tok/s131K ctx
dense
Granite Code 20B20B35.5 GB280 tok/s8K ctx
dense
Ministral 3 8B8B27.2 GB112 tok/s262K ctx
multimodal
MiniCPM-V 2.6 8B8B26.9 GB112 tok/s2K ctx
dense
Gemma 3 12B12B32.3 GB168 tok/s131K ctx
dense
Qwen 2.5 Coder 32B32B43.5 GB372 tok/s131K ctx
dense
Qwen 3 4B4B24.7 GB56 tok/s33K ctx
dense
Nemotron Nano 9B v29B28.0 GB126 tok/s131K ctx
dense
SQLCoder 7B7B26.3 GB98 tok/s8K ctx
dense
Qwen 2.5 VL 7B7B25.2 GB98 tok/s33K ctx
dense
DeepSeek Coder V2 16B16B33.2 GB1639 tok/s131K ctx
moe
Granite Code 34B34B44.5 GB351 tok/s8K ctx
dense
Jina Embeddings v30.57B23.2 GB8 tok/s8K ctx
dense
Command R 35B35B43.9 GB342 tok/s131K ctx
dense
Magistral 7B7B26.3 GB98 tok/s8K ctx
dense
CodeGeeX 4 9B9B26.2 GB126 tok/s131K ctx
dense
BGE M30.57B22.4 GB8 tok/s8K ctx
dense
OLMo 2 13B13B30.5 GB182 tok/s33K ctx
dense
Gemma 4 E4B8B26.3 GB112 tok/s128K ctx
dense
DeepSeek R1 Distill 32B32B43.5 GB372 tok/s33K ctx
dense
StarCoder 15B15B45.5 GB210 tok/s8K ctx
dense
CodeLlama 13B Instruct13B40.2 GB182 tok/s16K ctx
dense
Nemotron 70B70B67.7 GB171 tok/s131K ctx
dense
Qwen 2.5 7B7B25.2 GB98 tok/s131K ctx
dense
Granite Code 8B8B26.9 GB112 tok/s8K ctx
dense
B
B Tier -- Good Fit (65 models)
Solid performance, may need quantization or reduced context.
DevStral 7B7B26.3 GB98 tok/s8K ctx
dense
DeepSeek R1 Distill 14B14B31.6 GB196 tok/s33K ctx
dense
Codestral Mamba 7B7B24.9 GB98 tok/s262K ctx
state-space
MPT-30B-Instruct30B65.1 GB317 tok/s8K ctx
dense
LLaVA 1.6 13B13B40.2 GB182 tok/s4K ctx
dense
mxbai Embed Large0.34B22.3 GB5 tok/s1K ctx
dense
Granite 4.1 8B8B27.4 GB112 tok/s131K ctx
dense
Command R+ 104B104B87.0 GB115 tok/s131K ctx
dense
Pixtral 12B12B29.9 GB168 tok/s131K ctx
dense
CodeLlama 7B Instruct7B32.2 GB98 tok/s16K ctx
dense
StarCoder 7B7B31.7 GB98 tok/s8K ctx
dense
Snowflake Arctic Embed L0.34B22.3 GB5 tok/s1K ctx
dense
Falcon 40B Instruct40B50.7 GB259 tok/s8K ctx
dense
Qwen 2.5 Coder 3B3B24.1 GB42 tok/s131K ctx
dense
Mixtral 8x22B141B109.5 GB162 tok/s66K ctx
moe
Nous Dolphin 13B13B41.7 GB182 tok/s16K ctx
dense
WizardLM 13B13B40.2 GB182 tok/s8K ctx
dense
Vicuna 13B13B40.2 GB182 tok/s4K ctx
dense
Gemma 2 27B27B47.8 GB266 tok/s8K ctx
dense
Nous Hermes 1.09B37.8 GB126 tok/s16K ctx
dense
Gemma 4 E2B5.1B23.7 GB71 tok/s128K ctx
dense
BGE Large EN v1.50.34B22.3 GB5 tok/s1K ctx
dense
GLM-4 9B9B26.2 GB126 tok/s128K ctx
dense
InternLM 7B7B32.2 GB98 tok/s8K ctx
dense
InternLM Chat 7B7B32.2 GB98 tok/s8K ctx
dense
Llama 3.1 8B8B26.9 GB112 tok/s128K ctx
dense
Solar 7B7B27.3 GB98 tok/s8K ctx
dense
WizardMath 7B7B26.3 GB98 tok/s4K ctx
dense
OLMo 2 7B7B26.3 GB98 tok/s4K ctx
dense
Nomic Embed Text v1.50.14B20.9 GB2 tok/s8K ctx
dense
Qwen 2.5 Coder 7B7B25.2 GB98 tok/s131K ctx
dense
Ministral 3 3B3B22.7 GB42 tok/s262K ctx
multimodal
Gemma 3 4B4B24.6 GB56 tok/s128K ctx
dense
Phi 4 Mini 4B4B24.0 GB56 tok/s128K ctx
dense
LLaVA 1.5 7B7B32.2 GB98 tok/s4K ctx
dense
Mixtral 8x7B47B50.7 GB483 tok/s33K ctx
moe
Qwen 2.5 Math 72B72B68.9 GB166 tok/s4K ctx
dense
Qwen 3.5 2B2B23.0 GB28 tok/s131K ctx
dense
Baichuan 13B13B41.7 GB182 tok/s8K ctx
dense
Cerebras-GPT 13B13B39.2 GB182 tok/s131K ctx
dense
DeepSeek R1 Distill 7B7B25.2 GB98 tok/s33K ctx
dense
MPT-7B-Instruct7B32.2 GB98 tok/s8K ctx
dense
Baichuan 7B7B32.2 GB98 tok/s8K ctx
dense
DeepSeek R1 Distill 8B8B26.9 GB112 tok/s33K ctx
dense
Samantha 7B7B26.3 GB98 tok/s4K ctx
dense
Qwen 2.5 3B3B24.1 GB42 tok/s131K ctx
dense
Falcon 7B Instruct7B24.5 GB98 tok/s8K ctx
dense
Phi 3 Mini 3.8B3.8B28.3 GB53 tok/s128K ctx
dense
Qwen 3 1.7B1.7B22.8 GB24 tok/s33K ctx
dense
Gemma 2 9B9B30.7 GB126 tok/s8K ctx
dense
All MiniLM L6 v20.02B20.4 GB2 tok/s0K ctx
dense
Qwen 2.5 Coder 14B14B31.6 GB196 tok/s131K ctx
dense
Llama 3.2 11B Vision11B28.8 GB154 tok/s16K ctx
vision
Yi 1.5 34B34B44.5 GB352 tok/s4K ctx
dense
DeepSeek LLM 67B67B66.8 GB179 tok/s4K ctx
dense
Phi 3.5 Mini 4B4B28.4 GB56 tok/s128K ctx
dense
Granite 4.1 3B3B23.2 GB42 tok/s131K ctx
dense
Qwen 2.5 Coder 1.5B1.5B21.4 GB21 tok/s33K ctx
dense
Granite Code 3B3B24.4 GB42 tok/s8K ctx
dense
Mistral Nemo 12B12B29.9 GB168 tok/s128K ctx
dense
Yi Coder 9B9B27.1 GB126 tok/s131K ctx
dense
Mistral 7B Instruct v0.37B26.3 GB98 tok/s8K ctx
dense
Phi 3 Medium 14B14B31.7 GB196 tok/s128K ctx
dense
Codestral 22B22B36.0 GB308 tok/s33K ctx
dense
InternLM 20B20B52.8 GB280 tok/s8K ctx
dense
C
C Tier -- Tight Fit (191 models)
Runs with offloading or significant compromises.
Llama 3.2 3B3B23.6 GB42 tok/s128K ctx
dense
Ministral 8B8B27.2 GB112 tok/s131K ctx
dense
Qwen3.5 122B A10B122B94.2 GB105 tok/s125K ctx
dense
Solar Open 100B
C52100B92.8 GB110 tok/s151K ctx
dense
Solar Open 100B i1
C52100B92.8 GB110 tok/s151K ctx
dense
Aya Expanse 32B32B42.1 GB374 tok/s8K ctx
dense
K EXAONE 236B A23B
C51236B191.7 GB47 tok/s16K ctx
dense
Baichuan M3 235B
C51235B191.0 GB47 tok/s17K ctx
dense
Baichuan M3 235B i1
C51235B191.0 GB47 tok/s17K ctx
dense
DeepSeek R1 1.5B1.5B21.4 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B21.1 GB15 tok/s4K ctx
dense
Llama 3.3 70B Instruct
C5070B71.0 GB157 tok/s252K ctx
dense
SmolLM3 3B3B23.9 GB42 tok/s128K ctx
dense
Granite 3.1 8B8B26.9 GB112 tok/s128K ctx
state-space
Solar Open 69B REAP i1
C4969B70.3 GB160 tok/s257K ctx
dense
BaichuanMed OCR 72B i1
C4972B72.5 GB153 tok/s243K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C4970B71.0 GB157 tok/s252K ctx
dense
Yi 1.5 9B9B27.1 GB126 tok/s4K ctx
dense
Qwen 2.5 Math 7B7B25.2 GB98 tok/s4K ctx
dense
Yi 34B Chat34B44.5 GB352 tok/s200K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C4848B55.0 GB230 tok/s406K ctx
dense
Qwen 2.5 1.5B1.5B21.4 GB21 tok/s131K ctx
dense
Qwen3.5 35B A3B35B45.6 GB315 tok/s587K ctx
dense
Qwen3.5 35B A3B
C4735B45.6 GB315 tok/s587K ctx
dense
Hermes 4.3 36B
C4736B46.3 GB306 tok/s569K ctx
dense
StarCoder2 15B15B32.1 GB210 tok/s16K ctx
dense
OpenChat 7B7B26.3 GB98 tok/s8K ctx
dense
Qwen3.5 27B27B39.7 GB378 tok/s786K ctx
dense
gemma 3 27b it27B39.7 GB378 tok/s786K ctx
dense
EXAONE 4.0 32B
C4632B43.4 GB344 tok/s650K ctx
dense
aya expanse 32b heretic MPOA i1
C4632B43.4 GB344 tok/s650K ctx
dense
Aya Expanse 8B8B26.9 GB112 tok/s8K ctx
dense
Baichuan M2 32B Q4 K M
C4632B43.4 GB344 tok/s650K ctx
dense
gemma 3 27b it
C4627B39.7 GB378 tok/s786K ctx
dense
baichuan inc Baichuan M2 32B
C4632B43.4 GB344 tok/s650K ctx
dense
Mistral Small 3.2 24B Instruct 250624B37.6 GB336 tok/s895K ctx
dense
Gemma 3 1B1B21.1 GB14 tok/s33K ctx
dense
StableLM 2 12B12B40.9 GB168 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4624B37.6 GB336 tok/s895K ctx
dense
Mistral Small 24B Instruct 2501
C4624B37.6 GB336 tok/s895K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4624B37.6 GB336 tok/s895K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4624B37.6 GB336 tok/s895K ctx
dense
Vicuna 7B7B32.2 GB98 tok/s4K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4624B37.6 GB336 tok/s895K ctx
dense
Codestral 22B v0.1
C4622B36.1 GB308 tok/s984K ctx
dense
Codestral 22B v0.1
C4622B36.1 GB308 tok/s984K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4624B37.6 GB336 tok/s895K ctx
dense
Codestral 22B v0.1
C4622B36.1 GB308 tok/s984K ctx
dense
OpenHermes 2.5 7B7B26.3 GB98 tok/s8K ctx
dense
Dolphin 2.9 8B8B26.9 GB112 tok/s33K ctx
dense
Codestral 22B v0.1 IMat
C4622B36.1 GB308 tok/s984K ctx
dense
DeepSeek R1 Distill Qwen 14B14B30.3 GB196 tok/s1593K ctx
dense
Zephyr 7B Beta7B26.3 GB98 tok/s33K ctx
dense
Codestral 22B v0.1 i1
C4522B36.1 GB308 tok/s984K ctx
dense
internlm2 5 20b chat
C4520B34.6 GB280 tok/s1090K ctx
dense
Codestral 21B Pruned i1
C4521B35.4 GB294 tok/s1034K ctx
dense
Gemma 2 2B2B22.9 GB28 tok/s8K ctx
dense
internlm2 limarp chat 20b
C4520B34.6 GB280 tok/s1090K ctx
dense
Phi 4 reasoning vision 15B
C4515B31.0 GB210 tok/s1481K ctx
dense
internlm2 math plus 20b i1
C4520B34.6 GB280 tok/s1090K ctx
dense
StarCoder2 15B
C4515B31.0 GB210 tok/s1481K ctx
dense
Qwen3.5 9B9B26.6 GB126 tok/s2525K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C459B26.6 GB126 tok/s2525K ctx
dense
gemma 3 12b it
C4512B28.8 GB168 tok/s1873K ctx
dense
NousResearch Hermes 4 14B
C4514B30.3 GB196 tok/s1593K ctx
dense
Codestral RAG 19B Pruned i1
C4519B33.9 GB266 tok/s1152K ctx
dense
Meta Llama 3.1 8B Instruct
C458B25.9 GB112 tok/s2850K ctx
dense
Starling LM 7B7B26.3 GB98 tok/s8K ctx
dense
llava llama 3 8b v1 1
C458B25.9 GB112 tok/s2850K ctx
dense
Qwen3.5 9B
C459B26.6 GB126 tok/s2525K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B25.9 GB112 tok/s2850K ctx
dense
starcoder2 15b instruct v0.1
C4515B31.0 GB210 tok/s1481K ctx
dense
Qwen 2.5 Coder 0.5B0.5B20.6 GB7 tok/s131K ctx
dense
Llama 2 7B Chat
C457B25.2 GB98 tok/s3270K ctx
dense
starcoder2 15b i1
C4515B31.0 GB210 tok/s1481K ctx
dense
Meta Llama 3 8B Instruct
C458B25.9 GB112 tok/s2850K ctx
dense
DeepSeek R1 Distill Llama 8B8B25.9 GB112 tok/s2850K ctx
dense
starcoder2 15b instruct v0.1
C4515B31.0 GB210 tok/s1481K ctx
dense
Mistral 7B Instruct v0.2
C457B25.2 GB98 tok/s3270K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B25.9 GB112 tok/s2850K ctx
dense
Llama 3 8B Instruct 32k v0.1
C458B25.9 GB112 tok/s2850K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4510.7B27.9 GB150 tok/s2110K ctx
dense
Meta Llama 3.1 8B Instruct
C458B25.9 GB112 tok/s2850K ctx
dense
vntl llama3 8b v2
C458B25.9 GB112 tok/s2850K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4514B30.3 GB196 tok/s1593K ctx
dense
Mistral 7B Instruct v0.3
C457B25.2 GB98 tok/s3270K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4510.7B27.9 GB150 tok/s2110K ctx
dense
Yi Coder 9B Chat
C459B26.6 GB126 tok/s2525K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B25.9 GB112 tok/s2850K ctx
dense
glm 4 9b chat 1m
C459B26.6 GB126 tok/s2525K ctx
dense
HelpingAI 15B i1
C4515B31.0 GB210 tok/s1481K ctx
dense
internlm JanusCoder 14B
C4514B30.3 GB196 tok/s1593K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4514B30.3 GB196 tok/s1593K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B25.9 GB112 tok/s2850K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C458B25.9 GB112 tok/s2850K ctx
dense
Hermes 3 Llama 3.1 8B
C458B25.9 GB112 tok/s2850K ctx
dense
Yi 1.5 9B Chat
C459B26.6 GB126 tok/s2525K ctx
dense
Hermes 2 Pro Llama 3 8B
C458B25.9 GB112 tok/s2850K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B25.9 GB112 tok/s2850K ctx
dense
Mistral 7B Instruct v0.3
C457B25.2 GB98 tok/s3270K ctx
dense
Hermes 2 Pro Mistral 7B
C457B25.2 GB98 tok/s3270K ctx
dense
granite 8b code instruct 4k
C458B25.9 GB112 tok/s2850K ctx
dense
dolphin 2.9.4 llama3.1 8b
C458B25.9 GB112 tok/s2850K ctx
dense
zephyr 7B beta
C457B25.2 GB98 tok/s3270K ctx
dense
solar finalised finetuned Model 10.7B i1
C4510.7B27.9 GB150 tok/s2110K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C457B25.2 GB98 tok/s3270K ctx
dense
SOLAR 10.7B v1.0
C4510.7B27.9 GB150 tok/s2110K ctx
dense
openchat 3.6 8b 20240522 IMat
C458B25.9 GB112 tok/s2850K ctx
dense
Falcon H1R 7B
C457B25.2 GB98 tok/s3270K ctx
dense
aya expanse 8b
C458B25.9 GB112 tok/s2850K ctx
dense
EXAONE 3.5 7.8B Instruct
C457.8B25.8 GB109 tok/s2926K ctx
dense
falcon mamba 7b instruct Q4 K M
C457B25.2 GB98 tok/s3270K ctx
dense
zephyr 7B alpha
C457B25.2 GB98 tok/s3270K ctx
dense
dolphin v2 8b abliterated i1
C458B25.9 GB112 tok/s2850K ctx
dense
speechless zephyr code functionary 7b
C457B25.2 GB98 tok/s3270K ctx
dense
EXAONE 3.5 7.8B Instruct
C457.8B25.8 GB109 tok/s2926K ctx
dense
Yi 9B Coder i1
C449B26.6 GB126 tok/s2525K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C447.8B25.8 GB109 tok/s2926K ctx
dense
StarCoder2 7B
C447B25.2 GB98 tok/s3270K ctx
dense
aya expanse 8b orthogonal heretic i1
C448B25.9 GB112 tok/s2850K ctx
dense
exaone 3.0 7.8b it
C447.8B25.8 GB109 tok/s2926K ctx
dense
japanese stablelm instruct gamma 7B
C447B25.2 GB98 tok/s3270K ctx
dense
Falcon H1 7B Instruct
C447B25.2 GB98 tok/s3270K ctx
dense
HelpingAI2 9B
C449B26.6 GB126 tok/s2525K ctx
dense
Falcon H1R 7B7B25.2 GB98 tok/s3270K ctx
dense
Yi 1.5 6B Chat
C446B24.5 GB84 tok/s3828K ctx
dense
HelpingAI2.5 10B i1
C4410B27.4 GB140 tok/s2264K ctx
dense
Helply 10.2b chat i1
C4410.2B27.5 GB143 tok/s2218K ctx
dense
aya expanse 8b orthogonal heretic
C448B25.9 GB112 tok/s2850K ctx
dense
HelpingAI2 9B i1
C449B26.6 GB126 tok/s2525K ctx
dense
HelpingAI 9B 200k i1
C449B26.6 GB126 tok/s2525K ctx
dense
stablelm 2 zephyr 1 6b
C446B24.5 GB84 tok/s3828K ctx
dense
Mamba Codestral 7B v0.1
C447B25.2 GB98 tok/s3270K ctx
dense
stabilityai japanese stablelm base gamma 7b
C447B25.2 GB98 tok/s3270K ctx
dense
HelpingAI 9B i1
C449B26.6 GB126 tok/s2525K ctx
dense
baichuan2 7b chat
C447B25.2 GB98 tok/s3270K ctx
dense
DiscoPOP zephyr 7b gemma
C447B25.2 GB98 tok/s3270K ctx
dense
internlm2 5 1 8b chat i1
C448B25.9 GB112 tok/s2850K ctx
dense
internlm3 8b instruct abliterated i1
C448B25.9 GB112 tok/s2850K ctx
dense
starcoder2 7b
C447B25.2 GB98 tok/s3270K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C447B25.2 GB98 tok/s3270K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C447B25.2 GB98 tok/s3270K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C447B25.2 GB98 tok/s3270K ctx
dense
internlm2 math plus 7b IMat
C447B25.2 GB98 tok/s3270K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C447B25.2 GB98 tok/s3270K ctx
dense
internlm2 5 7b chat i1
C447B25.2 GB98 tok/s3270K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C447B25.2 GB98 tok/s3270K ctx
dense
MD Judge v0 2 internlm2 7b i1
C447B25.2 GB98 tok/s3270K ctx
dense
Yi 1.5 6B Chat
C446B24.5 GB84 tok/s3828K ctx
dense
Yi 1.5 6B6B24.7 GB84 tok/s4K ctx
dense
stablelm 2 1 6b chat imatrix
C446B24.5 GB84 tok/s3828K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C447B25.2 GB98 tok/s3270K ctx
dense
jointpreferences mistral 7b sft helpful
C447B25.2 GB98 tok/s3270K ctx
dense
zephyr 7b dpo full i1
C447B25.2 GB98 tok/s3270K ctx
dense
blossom v3 baichuan2 7b i1
C447B25.2 GB98 tok/s3270K ctx
dense
blossom v1 baichuan 7b i1
C447B25.2 GB98 tok/s3270K ctx
dense
Neural Chat 7B7B26.3 GB98 tok/s8K ctx
dense
HelpingAI2 6B
C446B24.5 GB84 tok/s3828K ctx
dense
StarCoder2 7B7B24.9 GB98 tok/s16K ctx
dense
HelpingAI2 6B i1
C446B24.5 GB84 tok/s3828K ctx
dense
Nemotron Mini 4B4B24.5 GB56 tok/s4K ctx
dense
DeepSeek LLM 7B7B31.7 GB98 tok/s4K ctx
dense
HelpingAI2.5 5B i1
C435B23.7 GB70 tok/s4611K ctx
dense
Qwen3.5 4B4B23.0 GB56 tok/s5784K ctx
dense
gemma 3 4b it
C434B23.0 GB56 tok/s5784K ctx
dense
gemma 3 4b it
C434B23.0 GB56 tok/s5784K ctx
dense
Qwen 3 0.6B0.6B21.3 GB8 tok/s33K ctx
dense
Llama 3.2 1B1B21.2 GB14 tok/s128K ctx
dense
Llama 3.2 3B Instruct
C423B22.6 GB42 tok/s7725K ctx
dense
Qwen2.5 3B Instruct3B22.3 GB42 tok/s7740K ctx
dense
Llama 3.2 3B Instruct
C423B22.3 GB42 tok/s7740K ctx
dense
Ministral 3 3B Instruct 2512
C423B22.3 GB42 tok/s7740K ctx
dense
HELVETE 3B
C423B22.3 GB42 tok/s7740K ctx
dense
Hermes 3 Llama 3.2 3B
C423B22.3 GB42 tok/s7740K ctx
dense
stablelm zephyr 3b
C423B22.3 GB42 tok/s7740K ctx
dense
StarCoder2 3B
C423B22.3 GB42 tok/s7740K ctx
dense
AI21 Jamba Reasoning 3B
C423B22.3 GB42 tok/s7740K ctx
dense
stablelm 3b 4e1t
C413B22.3 GB42 tok/s7740K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C413B22.3 GB42 tok/s7740K ctx
dense
ai21labs AI21 Jamba2 3B
C413B22.3 GB42 tok/s7740K ctx
dense
HelpingAI 3B hindi i1
C413B22.3 GB42 tok/s7740K ctx
dense
AI21 Jamba2 3B
C413B22.3 GB42 tok/s7740K ctx
dense
HelpingAI 3B hindi
C413B22.3 GB42 tok/s7740K ctx
dense
AI21 Jamba2 3B i1
C413B22.3 GB42 tok/s7740K ctx
dense
EXAONE 3.5 2.4B Instruct
C412.4B21.8 GB34 tok/s9696K ctx
dense
gemma 2 2b it
C412B22.0 GB28 tok/s11623K ctx
dense
gemma 2b2B21.6 GB28 tok/s11652K ctx
dense
Gemmasutra Mini 2B v1
C412B21.6 GB28 tok/s11652K ctx
dense
Qwen 3.5 0.6B0.6B21.3 GB8 tok/s131K ctx
dense
gemma 2 2b it
C412B21.6 GB28 tok/s11652K ctx
dense
StarCoder2 3B3B22.4 GB42 tok/s16K ctx
dense
F
F Tier -- Can't Run (36 models)
Too heavy for this hardware.
Qwen2.5 1.5B Instruct1.5B21.2 GB21 tok/s15563K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B21.2 GB21 tok/s15563K ctx
dense
Yi Coder 1.5B Chat
D401.5B21.2 GB21 tok/s15563K ctx
dense
stablelm 2 zephyr 1.6b
D391.6B21.3 GB22 tok/s14586K ctx
dense
logos16v2 stablelm2 1.6b i1
D391.6B21.3 GB22 tok/s14586K ctx
dense
Yi Coder 1.5B
D391.5B21.2 GB21 tok/s15563K ctx
dense
Falcon H1 1.5B Instruct1.5B21.2 GB21 tok/s15563K ctx
dense
TinyLlama 1.1B Chat v1.0
D391.1B20.9 GB15 tok/s21253K ctx
dense
Qwen 2.5 0.5B0.5B20.6 GB7 tok/s131K ctx
dense
EXAONE 4.0 1.2B
D391.2B21.0 GB17 tok/s19475K ctx
dense
Llama 3.2 1B Instruct Q8 0
D381B21.0 GB14 tok/s23358K ctx
dense
TinyLlama 1.1B Chat v0.3
D381.1B20.9 GB15 tok/s21253K ctx
dense
TinyLlama 1.1B Chat v0.61.1B20.9 GB15 tok/s21253K ctx
dense
Llama 3.2 1B Instruct
D381B20.8 GB14 tok/s23387K ctx
dense
gemma 3 1b it
D381B20.8 GB14 tok/s23387K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D381.1B20.9 GB15 tok/s21253K ctx
dense
Falcon3 1B Instruct abliterated
D381B20.8 GB14 tok/s23387K ctx
dense
SmolVLM 500M Instruct
D360.5B20.6 GB7 tok/s27438K ctx
dense
embeddinggemma 300M
D350.3B20.4 GB4 tok/s27465K ctx
dense
granite embedding 107m multilingual
D330.11B20.3 GB2 tok/s27494K ctx
dense
Falcon H1 Tiny 90M Instruct
D330.09B20.3 GB2 tok/s27495K ctx
dense
Qwen3-Coder 480B A35B Instruct480B315.8 GB26 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B265.1 GB46 tok/s4K ctx
moe
Mistral Large 3675B437.2 GB11 tok/s4K ctx
moe
GLM-5744B493.0 GB8 tok/s4K ctx
moe
GLM-5.1754B499.1 GB8 tok/s4K ctx
moe
Kimi K2.51000B637.5 GB6 tok/s4K ctx
moe
Kimi K2.61000B637.5 GB6 tok/s4K ctx
moe
DeepSeek V3 671B671B489.0 GB9 tok/s4K ctx
moe
DeepSeek V3.1 671B671B489.0 GB9 tok/s4K ctx
moe
DeepSeek V4 Pro1600B884.0 GB4 tok/s4K ctx
moe
DeepSeek R1 671B671B489.0 GB9 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B267.0 GB44 tok/s4K ctx
moe
Llama 3.1 405B405B274.8 GB14 tok/s4K ctx
dense
DeepSeek V3.2671B429.9 GB12 tok/s4K ctx
moe
Qwen3.5 397B A17B397B308.8 GB11 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs