AI Compatibility Tier List
AMD Instinct MI250 128GB
374 AI models ranked by how well they run on your 128 GB VRAM hardware. Best match: Qwen 3.5 122B A10B (score: 99).
374
Models tested
330
Can run
128 GB
VRAM
88
Best tok/s
Tier Distribution
S 28
A 54
B 60
C 188
F 44
S: 28A: 54B: 60C: 188F: 44
S
S Tier -- Perfect Fit (28 models)
These models run flawlessly with headroom to spare.
Qwen 3.5 122B A10B122B90.6 GB88 tok/s131K ctx
moe
Mistral Small 4 119B119B91.7 GB95 tok/s124K ctx
moe
Devstral 2 123B Instruct123B94.1 GB32 tok/s117K ctx
dense
GPT-OSS 120B117B90.0 GB33 tok/s131K ctx
dense
Command A 111B111B85.3 GB35 tok/s191K ctx
dense
Pixtral Large 124B124B94.7 GB31 tok/s115K ctx
dense
Leanstral 119B A6B119B95.1 GB87 tok/s76K ctx
moe
Qwen3-Coder-Next80B64.0 GB147 tok/s256K ctx
moe
Qwen 2.5 VL 72B72B62.5 GB54 tok/s33K ctx
dense
Qwen 3.6 35B A3B35B39.2 GB277 tok/s262K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B33.8 GB329 tok/s256K ctx
moe
Qwen 3.5 27B27B33.3 GB143 tok/s131K ctx
dense
Qwen3-VL 30B A3B Instruct30B33.5 GB340 tok/s256K ctx
moe
Qwen 3.5 35B A3B35B36.5 GB301 tok/s131K ctx
moe
Qwen 3.6 27B27B31.1 GB89 tok/s262K ctx
dense
Qwen 3 32B32B37.1 GB121 tok/s131K ctx
dense
Magistral Small 250724B30.8 GB160 tok/s131K ctx
dense
Devstral Small 2 24B Instruct24B30.8 GB160 tok/s256K ctx
dense
Qwen 3 30B A3B30.5B33.8 GB329 tok/s131K ctx
moe
Nemotron 3 Nano 30B30B34.4 GB128 tok/s131K ctx
dense
Qwen 3.5 9B9B21.4 GB126 tok/s131K ctx
dense
Qwen 3 14B14B24.7 GB196 tok/s131K ctx
dense
Devstral Small 1.124B30.8 GB160 tok/s131K ctx
dense
Gemma 4 31B30.7B47.1 GB76 tok/s104K ctx
dense
Phi-4-reasoning-plus 14B14.7B25.7 GB206 tok/s33K ctx
dense
Nemotron Cascade 2 30B A3B30B34.9 GB336 tok/s262K ctx
moe
Qwen 3 8B8B20.8 GB112 tok/s131K ctx
dense
GPT-OSS 20B21B29.0 GB418 tok/s128K ctx
moe
A
A Tier -- Great Fit (54 models)
Excellent performance with comfortable memory margins.
Llama 3.3 70B70B61.3 GB55 tok/s128K ctx
dense
Qwen 3.5 4B4B18.3 GB56 tok/s131K ctx
dense
Llama 4 Scout 17B 16E109B83.1 GB83 tok/s261K ctx
moe
EXAONE 4.0 32B32B37.1 GB120 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B32.7 GB353 tok/s256K ctx
moe
Qwen 2.5 72B72B62.5 GB54 tok/s131K ctx
dense
Llama 3.1 70B70B61.3 GB55 tok/s128K ctx
dense
Qwen 2.5 32B32B37.1 GB120 tok/s131K ctx
dense
Gemma 3 27B27B41.4 GB86 tok/s131K ctx
dense
Codestral 2 25.0822B29.6 GB164 tok/s256K ctx
dense
Ministral 3 14B14B24.7 GB196 tok/s262K ctx
multimodal
Mistral Small 3.2 24B24B30.8 GB160 tok/s131K ctx
vision
Kimi Linear 48B A3B48B43.9 GB74 tok/s1049K ctx
linear
OLMo 2 32B32B37.1 GB120 tok/s4K ctx
dense
Nemotron Nano 8B8B20.5 GB112 tok/s131K ctx
dense
LFM2 24B24B30.8 GB160 tok/s131K ctx
dense
Granite 4.1 30B30B35.9 GB128 tok/s131K ctx
dense
Phi-4 Mini Reasoning 4B3.8B17.5 GB53 tok/s131K ctx
dense
CogVLM2 19B19B27.7 GB202 tok/s8K ctx
dense
Mistral Small 24B24B30.8 GB160 tok/s33K ctx
dense
Mistral Small 3.1 24B24B30.8 GB160 tok/s131K ctx
dense
Phi-4 14B14B25.3 GB196 tok/s16K ctx
dense
InternVL2 8B8B20.5 GB112 tok/s8K ctx
dense
Qwen 2.5 14B14B25.2 GB196 tok/s131K ctx
dense
DeepSeek R1 Distill 70B70B61.3 GB55 tok/s131K ctx
dense
Granite Code 20B20B29.1 GB193 tok/s8K ctx
dense
Ministral 3 8B8B20.8 GB112 tok/s262K ctx
multimodal
Gemma 3 12B12B25.9 GB168 tok/s131K ctx
dense
Qwen 2.5 Coder 32B32B37.1 GB120 tok/s131K ctx
dense
MiniCPM-V 2.6 8B8B20.5 GB112 tok/s2K ctx
dense
Granite Code 34B34B38.1 GB114 tok/s8K ctx
dense
DeepSeek Coder V2 16B16B26.8 GB531 tok/s131K ctx
moe
Nemotron Nano 9B v29B21.6 GB126 tok/s131K ctx
dense
SQLCoder 7B7B19.9 GB98 tok/s8K ctx
dense
Qwen 3 4B4B18.3 GB56 tok/s33K ctx
dense
Qwen 2.5 VL 7B7B18.8 GB98 tok/s33K ctx
dense
Command R 35B35B37.5 GB111 tok/s131K ctx
dense
Jina Embeddings v30.57B16.8 GB8 tok/s8K ctx
dense
StarCoder 15B15B39.1 GB206 tok/s8K ctx
dense
DeepSeek R1 Distill 32B32B37.1 GB120 tok/s33K ctx
dense
Magistral 7B7B19.9 GB98 tok/s8K ctx
dense
CodeGeeX 4 9B9B19.8 GB126 tok/s131K ctx
dense
OLMo 2 13B13B24.1 GB182 tok/s33K ctx
dense
BGE M30.57B16.0 GB8 tok/s8K ctx
dense
CodeLlama 13B Instruct13B33.8 GB182 tok/s16K ctx
dense
Gemma 4 E4B8B19.9 GB112 tok/s128K ctx
dense
MPT-30B-Instruct30B58.7 GB103 tok/s8K ctx
dense
Nemotron 70B70B61.3 GB55 tok/s131K ctx
dense
Qwen 2.5 7B7B18.8 GB98 tok/s131K ctx
dense
LLaVA 1.6 13B13B33.8 GB182 tok/s4K ctx
dense
DeepSeek R1 Distill 14B14B25.2 GB196 tok/s33K ctx
dense
Granite Code 8B8B20.5 GB112 tok/s8K ctx
dense
DevStral 7B7B19.9 GB98 tok/s8K ctx
dense
Codestral Mamba 7B7B18.5 GB98 tok/s262K ctx
state-space
B
B Tier -- Good Fit (60 models)
Solid performance, may need quantization or reduced context.
Command R+ 104B104B80.6 GB37 tok/s131K ctx
dense
Falcon 40B Instruct40B44.3 GB84 tok/s8K ctx
dense
Granite 4.1 8B8B21.0 GB112 tok/s131K ctx
dense
CodeLlama 7B Instruct7B25.8 GB98 tok/s16K ctx
dense
mxbai Embed Large0.34B15.9 GB5 tok/s1K ctx
dense
StarCoder 7B7B25.3 GB98 tok/s8K ctx
dense
Pixtral 12B12B23.5 GB168 tok/s131K ctx
dense
Mixtral 8x22B141B103.1 GB53 tok/s66K ctx
moe
Snowflake Arctic Embed L0.34B15.9 GB5 tok/s1K ctx
dense
Nous Dolphin 13B13B35.3 GB182 tok/s16K ctx
dense
Gemma 2 27B27B41.4 GB86 tok/s8K ctx
dense
WizardLM 13B13B33.8 GB182 tok/s8K ctx
dense
Vicuna 13B13B33.8 GB182 tok/s4K ctx
dense
Qwen 2.5 Coder 3B3B17.7 GB42 tok/s131K ctx
dense
Nous Hermes 1.09B31.4 GB126 tok/s16K ctx
dense
Gemma 4 E2B5.1B17.3 GB71 tok/s128K ctx
dense
InternLM 7B7B25.8 GB98 tok/s8K ctx
dense
InternLM Chat 7B7B25.8 GB98 tok/s8K ctx
dense
BGE Large EN v1.50.34B15.9 GB5 tok/s1K ctx
dense
GLM-4 9B9B19.8 GB126 tok/s128K ctx
dense
Llama 3.1 8B8B20.5 GB112 tok/s128K ctx
dense
Solar 7B7B20.9 GB98 tok/s8K ctx
dense
WizardMath 7B7B19.9 GB98 tok/s4K ctx
dense
OLMo 2 7B7B19.9 GB98 tok/s4K ctx
dense
Nomic Embed Text v1.50.14B14.5 GB2 tok/s8K ctx
dense
Qwen 2.5 Coder 7B7B18.8 GB98 tok/s131K ctx
dense
Mixtral 8x7B47B44.3 GB157 tok/s33K ctx
moe
Gemma 3 4B4B18.2 GB56 tok/s128K ctx
dense
Ministral 3 3B3B16.3 GB42 tok/s262K ctx
multimodal
LLaVA 1.5 7B7B25.8 GB98 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B62.5 GB54 tok/s4K ctx
dense
Baichuan 13B13B35.3 GB182 tok/s8K ctx
dense
Phi 4 Mini 4B4B17.6 GB56 tok/s128K ctx
dense
Cerebras-GPT 13B13B32.8 GB182 tok/s131K ctx
dense
Qwen 3.5 2B2B16.6 GB28 tok/s131K ctx
dense
MPT-7B-Instruct7B25.8 GB98 tok/s8K ctx
dense
Baichuan 7B7B25.8 GB98 tok/s8K ctx
dense
DeepSeek R1 Distill 7B7B18.8 GB98 tok/s33K ctx
dense
DeepSeek R1 Distill 8B8B20.5 GB112 tok/s33K ctx
dense
Samantha 7B7B19.9 GB98 tok/s4K ctx
dense
Qwen 2.5 3B3B17.7 GB42 tok/s131K ctx
dense
Falcon 7B Instruct7B18.1 GB98 tok/s8K ctx
dense
Yi 1.5 34B34B38.1 GB114 tok/s4K ctx
dense
Phi 3 Mini 3.8B3.8B21.9 GB53 tok/s128K ctx
dense
Gemma 2 9B9B24.3 GB126 tok/s8K ctx
dense
Qwen 2.5 Coder 14B14B25.2 GB196 tok/s131K ctx
dense
DeepSeek LLM 67B67B60.4 GB58 tok/s4K ctx
dense
Qwen 3 1.7B1.7B16.4 GB24 tok/s33K ctx
dense
Llama 3.2 11B Vision11B22.4 GB154 tok/s16K ctx
vision
All MiniLM L6 v20.02B14.0 GB2 tok/s0K ctx
dense
Phi 3.5 Mini 4B4B22.0 GB56 tok/s128K ctx
dense
Granite 4.1 3B3B16.8 GB42 tok/s131K ctx
dense
Granite Code 3B3B18.0 GB42 tok/s8K ctx
dense
Qwen 2.5 Coder 1.5B1.5B15.0 GB21 tok/s33K ctx
dense
Mistral Nemo 12B12B23.5 GB168 tok/s128K ctx
dense
InternLM 20B20B46.4 GB154 tok/s8K ctx
dense
Yi Coder 9B9B20.7 GB126 tok/s131K ctx
dense
Mistral 7B Instruct v0.37B19.9 GB98 tok/s8K ctx
dense
Phi 3 Medium 14B14B25.3 GB196 tok/s128K ctx
dense
Codestral 22B22B29.6 GB174 tok/s33K ctx
dense
C
C Tier -- Tight Fit (188 models)
Runs with offloading or significant compromises.
Ministral 8B8B20.8 GB112 tok/s131K ctx
dense
Llama 3.2 3B3B17.2 GB42 tok/s128K ctx
dense
Qwen3.5 122B A10B122B87.8 GB34 tok/s61K ctx
dense
Aya Expanse 32B32B35.7 GB121 tok/s8K ctx
dense
Solar Open 100B
C53100B86.4 GB36 tok/s73K ctx
dense
Solar Open 100B i1
C53100B86.4 GB36 tok/s73K ctx
dense
Llama 3.3 70B Instruct
C5170B64.6 GB51 tok/s140K ctx
dense
Solar Open 69B REAP i1
C5069B63.9 GB52 tok/s143K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C5070B64.6 GB51 tok/s140K ctx
dense
BaichuanMed OCR 72B i1
C5072B66.1 GB50 tok/s133K ctx
dense
DeepSeek R1 1.5B1.5B15.0 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B14.7 GB15 tok/s4K ctx
dense
SmolLM3 3B3B17.5 GB42 tok/s128K ctx
dense
Granite 3.1 8B8B20.5 GB112 tok/s128K ctx
state-space
Yi 1.5 9B9B20.7 GB126 tok/s4K ctx
dense
Yi 34B Chat34B38.1 GB114 tok/s200K ctx
dense
Qwen 2.5 Math 7B7B18.8 GB98 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C4948B48.6 GB74 tok/s242K ctx
dense
Qwen3.5 35B A3B35B39.2 GB102 tok/s363K ctx
dense
Qwen3.5 35B A3B
C4935B39.2 GB102 tok/s363K ctx
dense
Hermes 4.3 36B
C4836B39.9 GB99 tok/s350K ctx
dense
EXAONE 4.0 32B
C4832B37.0 GB112 tok/s404K ctx
dense
Qwen3.5 27B27B33.3 GB132 tok/s495K ctx
dense
aya expanse 32b heretic MPOA i1
C4832B37.0 GB112 tok/s404K ctx
dense
Qwen 2.5 1.5B1.5B15.0 GB21 tok/s131K ctx
dense
Baichuan M2 32B Q4 K M
C4832B37.0 GB112 tok/s404K ctx
dense
gemma 3 27b it27B33.3 GB132 tok/s495K ctx
dense
baichuan inc Baichuan M2 32B
C4832B37.0 GB112 tok/s404K ctx
dense
StarCoder2 15B15B25.7 GB210 tok/s16K ctx
dense
gemma 3 27b it
C4827B33.3 GB132 tok/s495K ctx
dense
StableLM 2 12B12B34.5 GB168 tok/s4K ctx
dense
Mistral Small 3.2 24B Instruct 250624B31.2 GB149 tok/s567K ctx
dense
OpenChat 7B7B19.9 GB98 tok/s8K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4724B31.2 GB149 tok/s567K ctx
dense
Mistral Small 24B Instruct 2501
C4724B31.2 GB149 tok/s567K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4724B31.2 GB149 tok/s567K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4724B31.2 GB149 tok/s567K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4724B31.2 GB149 tok/s567K ctx
dense
Aya Expanse 8B8B20.5 GB112 tok/s8K ctx
dense
Codestral 22B v0.1
C4722B29.7 GB162 tok/s626K ctx
dense
Codestral 22B v0.1
C4722B29.7 GB162 tok/s626K ctx
dense
Vicuna 7B7B25.8 GB98 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4724B31.2 GB149 tok/s567K ctx
dense
Codestral 22B v0.1
C4722B29.7 GB162 tok/s626K ctx
dense
Codestral 22B v0.1 IMat
C4622B29.7 GB162 tok/s626K ctx
dense
Codestral 22B v0.1 i1
C4622B29.7 GB162 tok/s626K ctx
dense
Gemma 3 1B1B14.7 GB14 tok/s33K ctx
dense
Codestral 21B Pruned i1
C4621B29.0 GB170 tok/s660K ctx
dense
internlm2 5 20b chat
C4620B28.2 GB178 tok/s697K ctx
dense
internlm2 limarp chat 20b
C4620B28.2 GB178 tok/s697K ctx
dense
internlm2 math plus 20b i1
C4620B28.2 GB178 tok/s697K ctx
dense
DeepSeek R1 Distill Qwen 14B14B23.9 GB196 tok/s1031K ctx
dense
OpenHermes 2.5 7B7B19.9 GB98 tok/s8K ctx
dense
Dolphin 2.9 8B8B20.5 GB112 tok/s33K ctx
dense
Codestral RAG 19B Pruned i1
C4619B27.5 GB188 tok/s738K ctx
dense
Phi 4 reasoning vision 15B
C4615B24.6 GB210 tok/s957K ctx
dense
StarCoder2 15B
C4615B24.6 GB210 tok/s957K ctx
dense
Zephyr 7B Beta7B19.9 GB98 tok/s33K ctx
dense
NousResearch Hermes 4 14B
C4614B23.9 GB196 tok/s1031K ctx
dense
gemma 3 12b it
C4612B22.4 GB168 tok/s1217K ctx
dense
starcoder2 15b instruct v0.1
C4615B24.6 GB210 tok/s957K ctx
dense
starcoder2 15b i1
C4615B24.6 GB210 tok/s957K ctx
dense
Qwen3.5 9B9B20.2 GB126 tok/s1651K ctx
dense
starcoder2 15b instruct v0.1
C4615B24.6 GB210 tok/s957K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C469B20.2 GB126 tok/s1651K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4614B23.9 GB196 tok/s1031K ctx
dense
HelpingAI 15B i1
C4615B24.6 GB210 tok/s957K ctx
dense
Gemma 2 2B2B16.5 GB28 tok/s8K ctx
dense
Starling LM 7B7B19.9 GB98 tok/s8K ctx
dense
internlm JanusCoder 14B
C4514B23.9 GB196 tok/s1031K ctx
dense
Qwen3.5 9B
C459B20.2 GB126 tok/s1651K ctx
dense
Meta Llama 3.1 8B Instruct
C458B19.5 GB112 tok/s1867K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4514B23.9 GB196 tok/s1031K ctx
dense
llava llama 3 8b v1 1
C458B19.5 GB112 tok/s1867K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4510.7B21.5 GB150 tok/s1375K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B19.5 GB112 tok/s1867K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4510.7B21.5 GB150 tok/s1375K ctx
dense
Meta Llama 3 8B Instruct
C458B19.5 GB112 tok/s1867K ctx
dense
DeepSeek R1 Distill Llama 8B8B19.5 GB112 tok/s1867K ctx
dense
Llama 2 7B Chat
C457B18.8 GB98 tok/s2146K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B19.5 GB112 tok/s1867K ctx
dense
Llama 3 8B Instruct 32k v0.1
C458B19.5 GB112 tok/s1867K ctx
dense
Yi Coder 9B Chat
C459B20.2 GB126 tok/s1651K ctx
dense
glm 4 9b chat 1m
C459B20.2 GB126 tok/s1651K ctx
dense
Meta Llama 3.1 8B Instruct
C458B19.5 GB112 tok/s1867K ctx
dense
vntl llama3 8b v2
C458B19.5 GB112 tok/s1867K ctx
dense
Mistral 7B Instruct v0.2
C457B18.8 GB98 tok/s2146K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B19.5 GB112 tok/s1867K ctx
dense
Mistral 7B Instruct v0.3
C457B18.8 GB98 tok/s2146K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B19.5 GB112 tok/s1867K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C458B19.5 GB112 tok/s1867K ctx
dense
Hermes 3 Llama 3.1 8B
C458B19.5 GB112 tok/s1867K ctx
dense
Yi 1.5 9B Chat
C459B20.2 GB126 tok/s1651K ctx
dense
solar finalised finetuned Model 10.7B i1
C4510.7B21.5 GB150 tok/s1375K ctx
dense
Qwen 2.5 Coder 0.5B0.5B14.2 GB7 tok/s131K ctx
dense
Hermes 2 Pro Llama 3 8B
C458B19.5 GB112 tok/s1867K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B19.5 GB112 tok/s1867K ctx
dense
SOLAR 10.7B v1.0
C4510.7B21.5 GB150 tok/s1375K ctx
dense
granite 8b code instruct 4k
C458B19.5 GB112 tok/s1867K ctx
dense
dolphin 2.9.4 llama3.1 8b
C458B19.5 GB112 tok/s1867K ctx
dense
Mistral 7B Instruct v0.3
C457B18.8 GB98 tok/s2146K ctx
dense
Hermes 2 Pro Mistral 7B
C457B18.8 GB98 tok/s2146K ctx
dense
zephyr 7B beta
C457B18.8 GB98 tok/s2146K ctx
dense
openchat 3.6 8b 20240522 IMat
C458B19.5 GB112 tok/s1867K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C457B18.8 GB98 tok/s2146K ctx
dense
aya expanse 8b
C458B19.5 GB112 tok/s1867K ctx
dense
EXAONE 3.5 7.8B Instruct
C457.8B19.4 GB109 tok/s1917K ctx
dense
Falcon H1R 7B
C457B18.8 GB98 tok/s2146K ctx
dense
Yi 9B Coder i1
C459B20.2 GB126 tok/s1651K ctx
dense
dolphin v2 8b abliterated i1
C458B19.5 GB112 tok/s1867K ctx
dense
EXAONE 3.5 7.8B Instruct
C457.8B19.4 GB109 tok/s1917K ctx
dense
falcon mamba 7b instruct Q4 K M
C457B18.8 GB98 tok/s2146K ctx
dense
HelpingAI2.5 10B i1
C4510B21.0 GB140 tok/s1477K ctx
dense
Helply 10.2b chat i1
C4510.2B21.1 GB143 tok/s1447K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C457.8B19.4 GB109 tok/s1917K ctx
dense
zephyr 7B alpha
C457B18.8 GB98 tok/s2146K ctx
dense
speechless zephyr code functionary 7b
C457B18.8 GB98 tok/s2146K ctx
dense
aya expanse 8b orthogonal heretic i1
C458B19.5 GB112 tok/s1867K ctx
dense
HelpingAI2 9B
C459B20.2 GB126 tok/s1651K ctx
dense
exaone 3.0 7.8b it
C457.8B19.4 GB109 tok/s1917K ctx
dense
StarCoder2 7B
C457B18.8 GB98 tok/s2146K ctx
dense
japanese stablelm instruct gamma 7B
C457B18.8 GB98 tok/s2146K ctx
dense
HelpingAI2 9B i1
C459B20.2 GB126 tok/s1651K ctx
dense
Falcon H1 7B Instruct
C457B18.8 GB98 tok/s2146K ctx
dense
HelpingAI 9B 200k i1
C459B20.2 GB126 tok/s1651K ctx
dense
Falcon H1R 7B7B18.8 GB98 tok/s2146K ctx
dense
aya expanse 8b orthogonal heretic
C458B19.5 GB112 tok/s1867K ctx
dense
HelpingAI 9B i1
C459B20.2 GB126 tok/s1651K ctx
dense
Yi 1.5 6B Chat
C456B18.1 GB84 tok/s2518K ctx
dense
Mamba Codestral 7B v0.1
C457B18.8 GB98 tok/s2146K ctx
dense
internlm2 5 1 8b chat i1
C458B19.5 GB112 tok/s1867K ctx
dense
stabilityai japanese stablelm base gamma 7b
C457B18.8 GB98 tok/s2146K ctx
dense
internlm3 8b instruct abliterated i1
C458B19.5 GB112 tok/s1867K ctx
dense
stablelm 2 zephyr 1 6b
C456B18.1 GB84 tok/s2518K ctx
dense
baichuan2 7b chat
C457B18.8 GB98 tok/s2146K ctx
dense
DiscoPOP zephyr 7b gemma
C457B18.8 GB98 tok/s2146K ctx
dense
starcoder2 7b
C457B18.8 GB98 tok/s2146K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C457B18.8 GB98 tok/s2146K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C457B18.8 GB98 tok/s2146K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C457B18.8 GB98 tok/s2146K ctx
dense
internlm2 math plus 7b IMat
C457B18.8 GB98 tok/s2146K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C457B18.8 GB98 tok/s2146K ctx
dense
internlm2 5 7b chat i1
C457B18.8 GB98 tok/s2146K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C457B18.8 GB98 tok/s2146K ctx
dense
MD Judge v0 2 internlm2 7b i1
C457B18.8 GB98 tok/s2146K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C447B18.8 GB98 tok/s2146K ctx
dense
jointpreferences mistral 7b sft helpful
C447B18.8 GB98 tok/s2146K ctx
dense
zephyr 7b dpo full i1
C447B18.8 GB98 tok/s2146K ctx
dense
blossom v3 baichuan2 7b i1
C447B18.8 GB98 tok/s2146K ctx
dense
Yi 1.5 6B6B18.3 GB84 tok/s4K ctx
dense
Yi 1.5 6B Chat
C446B18.1 GB84 tok/s2518K ctx
dense
DeepSeek LLM 7B7B25.3 GB98 tok/s4K ctx
dense
Neural Chat 7B7B19.9 GB98 tok/s8K ctx
dense
blossom v1 baichuan 7b i1
C447B18.8 GB98 tok/s2146K ctx
dense
stablelm 2 1 6b chat imatrix
C446B18.1 GB84 tok/s2518K ctx
dense
HelpingAI2 6B
C446B18.1 GB84 tok/s2518K ctx
dense
StarCoder2 7B7B18.5 GB98 tok/s16K ctx
dense
HelpingAI2 6B i1
C446B18.1 GB84 tok/s2518K ctx
dense
Nemotron Mini 4B4B18.1 GB56 tok/s4K ctx
dense
HelpingAI2.5 5B i1
C435B17.3 GB70 tok/s3038K ctx
dense
Qwen3.5 4B4B16.6 GB56 tok/s3818K ctx
dense
gemma 3 4b it
C434B16.6 GB56 tok/s3818K ctx
dense
gemma 3 4b it
C434B16.6 GB56 tok/s3818K ctx
dense
Qwen 3 0.6B0.6B14.9 GB8 tok/s33K ctx
dense
Llama 3.2 1B1B14.8 GB14 tok/s128K ctx
dense
Llama 3.2 3B Instruct
C423B16.2 GB42 tok/s5104K ctx
dense
Qwen2.5 3B Instruct3B15.9 GB42 tok/s5119K ctx
dense
Llama 3.2 3B Instruct
C423B15.9 GB42 tok/s5119K ctx
dense
Ministral 3 3B Instruct 2512
C423B15.9 GB42 tok/s5119K ctx
dense
HELVETE 3B
C423B15.9 GB42 tok/s5119K ctx
dense
Hermes 3 Llama 3.2 3B
C423B15.9 GB42 tok/s5119K ctx
dense
stablelm zephyr 3b
C423B15.9 GB42 tok/s5119K ctx
dense
StarCoder2 3B
C423B15.9 GB42 tok/s5119K ctx
dense
AI21 Jamba Reasoning 3B
C423B15.9 GB42 tok/s5119K ctx
dense
stablelm 3b 4e1t
C423B15.9 GB42 tok/s5119K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C423B15.9 GB42 tok/s5119K ctx
dense
ai21labs AI21 Jamba2 3B
C423B15.9 GB42 tok/s5119K ctx
dense
HelpingAI 3B hindi i1
C423B15.9 GB42 tok/s5119K ctx
dense
AI21 Jamba2 3B
C423B15.9 GB42 tok/s5119K ctx
dense
HelpingAI 3B hindi
C413B15.9 GB42 tok/s5119K ctx
dense
AI21 Jamba2 3B i1
C413B15.9 GB42 tok/s5119K ctx
dense
EXAONE 3.5 2.4B Instruct
C412.4B15.4 GB34 tok/s6419K ctx
dense
gemma 2 2b it
C412B15.6 GB28 tok/s7691K ctx
dense
gemma 2b2B15.2 GB28 tok/s7720K ctx
dense
Gemmasutra Mini 2B v1
C412B15.2 GB28 tok/s7720K ctx
dense
Qwen 3.5 0.6B0.6B14.9 GB8 tok/s131K ctx
dense
gemma 2 2b it
C412B15.2 GB28 tok/s7720K ctx
dense
StarCoder2 3B3B16.0 GB42 tok/s16K ctx
dense
F
F Tier -- Can't Run (44 models)
Too heavy for this hardware.
Qwen2.5 1.5B Instruct1.5B14.8 GB21 tok/s10321K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B14.8 GB21 tok/s10321K ctx
dense
Yi Coder 1.5B Chat
D401.5B14.8 GB21 tok/s10321K ctx
dense
stablelm 2 zephyr 1.6b
D391.6B14.9 GB22 tok/s9670K ctx
dense
logos16v2 stablelm2 1.6b i1
D391.6B14.9 GB22 tok/s9670K ctx
dense
Yi Coder 1.5B
D391.5B14.8 GB21 tok/s10321K ctx
dense
Falcon H1 1.5B Instruct1.5B14.8 GB21 tok/s10321K ctx
dense
TinyLlama 1.1B Chat v1.0
D391.1B14.5 GB15 tok/s14104K ctx
dense
Qwen 2.5 0.5B0.5B14.2 GB7 tok/s131K ctx
dense
EXAONE 4.0 1.2B
D391.2B14.6 GB17 tok/s12922K ctx
dense
Llama 3.2 1B Instruct Q8 0
D391B14.6 GB14 tok/s15494K ctx
dense
TinyLlama 1.1B Chat v0.3
D391.1B14.5 GB15 tok/s14104K ctx
dense
TinyLlama 1.1B Chat v0.61.1B14.5 GB15 tok/s14104K ctx
dense
Llama 3.2 1B Instruct
D381B14.4 GB14 tok/s15522K ctx
dense
gemma 3 1b it
D381B14.4 GB14 tok/s15522K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D381.1B14.5 GB15 tok/s14104K ctx
dense
Falcon3 1B Instruct abliterated
D381B14.4 GB14 tok/s15522K ctx
dense
SmolVLM 500M Instruct
D360.5B14.2 GB7 tok/s18222K ctx
dense
embeddinggemma 300M
D350.3B14.0 GB4 tok/s18249K ctx
dense
granite embedding 107m multilingual
D330.11B13.9 GB2 tok/s18278K ctx
dense
Falcon H1 Tiny 90M Instruct
D330.09B13.9 GB2 tok/s18279K ctx
dense
Qwen 3 235B A22B235B159.9 GB21 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B309.4 GB3 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B258.7 GB5 tok/s4K ctx
moe
Mistral Large 3675B430.8 GB3 tok/s4K ctx
moe
GLM-5744B486.6 GB2 tok/s4K ctx
moe
GLM-5.1754B492.7 GB2 tok/s4K ctx
moe
Kimi K2.51000B631.1 GB2 tok/s4K ctx
moe
Kimi K2.61000B631.1 GB2 tok/s4K ctx
moe
DeepSeek V3 671B671B482.6 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B482.6 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B877.6 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B173.0 GB19 tok/s4K ctx
moe
DeepSeek R1 671B671B482.6 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B216.3 GB10 tok/s4K ctx
moe
DeepSeek V2.5 236B236B216.3 GB10 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B260.6 GB5 tok/s4K ctx
moe
Llama 3.1 405B405B268.4 GB2 tok/s4K ctx
dense
DeepSeek V3.2671B423.5 GB3 tok/s4K ctx
moe
MiniMax M2.7
F0230B157.8 GB24 tok/s4K ctx
moe
K EXAONE 236B A23B
F0236B185.3 GB5 tok/s4K ctx
dense
Baichuan M3 235B
F0235B184.6 GB5 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B184.6 GB5 tok/s4K ctx
dense
Qwen3.5 397B A17B397B302.4 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs